University of Southern California Dissertations and Theses

Targeted cell imaging by in-situ assembly and activation of hot spot SERS nanoprobes using split-fluorescent protein scaffolds

(USC Thesis Other)

Targeted cell imaging by in-situ assembly and activation of hot spot SERS nanoprobes using split-fluorescent protein scaffolds

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

TARGETED CELL IMAGING BY IN-SITU ASSEMBLY AND ACTIVATION OF HOT SPOT
SERS NANOPROBES USING SPLIT-FLUORESCENT PROTEIN SCAFFOLDS

by

Tuğba Kӧker

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(MOLECULAR BIOLOGY)

August 2017

Copyright 2017 Tuğba Kӧker

ii

DEDICATION

To the human knowledge…

iii

ACKNOWLEDGEMENTS
First of all, I would like to thank my advisor, Fabien Pinaud. I sincerely appreciate for his
guidance and support from day one. I have learned so many different skills and have become a
detail-oriented scientist. It was an honor to work with him, and I will always remember and be
thankful for his effort on my doctorate journey.
Without fellow researchers, it would be hard to finish a dissertation. Pinaud Group
members, I would like thank every single one of you for listening to me as I was explaining my
experiments without any early warning about how long it would take. Ram, it was so much fun to
start together in this lab, there was not even a single day without saying our names out loud or if I
quote “There is so much stuff to do”. Tony, I appreciate all the single molecule work that you did
for me. I will miss those moments that you ran around saying “We have contamination
everywhere”, and Tony, we do not have contamination! Markville, you should have come earlier,
you are my soul lab mate. I enjoyed all the time we spent together in the lab. Next time (!) I’d
love to work with you on the same project. Yunke, you are the happiest lab mate ever, thank you
for cheering up the lab in my last year. I will never forget your smile and optimism. Taerin, I’m
sincerely happy for working and collaborating with you on the same project. I will always
remember your understanding and calmness.
I have to mention about a very special group of people that made my PhD life vibrant.
People of Normandie House, we shared this 6 years all together with lots of laughter, food,
drinks, long discussions, and friendship. Eliseo, Brad, Matilde, Amanda, Ulrike, Thomas, and
Wendy, I will always remember all the great memories we have created.
My dearest parents, Hatice and Ertuğrul, you showed this incredible support throughout
my life, for every decision I have made. I truly cannot express my gratitude enough for this
endless support and understanding. Thank you so much for being there for me, regardless.
iv

My beloved boyfriend, fiancé, and soon to be husband, Mehmet, I do not know if there is
any word that is powerful enough to express my feelings, my gratitude, and my admiration to
you. You acknowledged me on your own PhD thesis, and said that you hoped one day you could
help me the same way. You have always been there for me from day one in this period of my life,
long discussions on my experiments, sleepless nights while I was in the lab, endless support, 5
years of back and forth flights, and yet amazing time and emotions. I will never forget your trust
in my abilities, and I will always remember your excitement about my achievements. You will
always hold a special place in my heart.

v

TABLE OF CONTENTS
Fundamentals of Raman Spectroscopy .......................................................................... 1
Introduction .............................................................................................................. 1
Physical Model of Raman Spectroscopy .................................................................. 3
Resonance Raman Spectroscopy and Surface Enhanced Raman
Spectroscopy ............................................................................................................ 8
Plasmonic materials ................................................................................................ 10
Conclusion .............................................................................................................. 16
Surface Chemistry of Metal Nanoparticles .................................................................. 18
1.2.1 Introduction ............................................................................................................ 18
1.2.2 Surface modifications ............................................................................................. 20
1.2.3 Poly(ethylene) Glycol (PEG) coating ..................................................................... 23
1.2.4 Raman reporters ..................................................................................................... 25
1.2.5 Hot spot formation of metal nanoparticles ............................................................. 29
1.2.6 Conclusion .............................................................................................................. 33
Biomedical Applications of Sers Probes ..................................................................... 35
1.3.1 Introduction ............................................................................................................ 35
1.3.2 SERS nanoprobes for biosensing ........................................................................... 36
1.3.3 SERS nanoprobes for biomedical imaging ............................................................. 38
1.3.4 SERS nanoprobes for therapeutic and theranostic applications ............................. 47
1.3.5 Conclusion .............................................................................................................. 53
Introduction ................................................................................................................. 56
GFP structure and self-assembly of split-GFP ............................................................ 58
Point mutations in split-GFP to generate split-GFP variants ....................................... 61
Point mutations in split-GFP to generate split-YFP and split-CFP spectral variants .. 62
Photophysical characteristics of full-length FP and split-FP variants ......................... 65
Optical spectra of full-length FP and split-FP variants .......................................... 65
Fluorescence lifetime of full-length FP and split-FP variants ................................ 66
DEDICATION ................................................................................................................................. ii
ACKNOWLEDGEMENTS ............................................................................................................ iii
TABLE OF CONTENTS ................................................................................................................. v
LIST OF FIGURES ...................................................................................................................... viii
LIST OF TABLES ........................................................................................................................ xxi
ABSTRACT ................................................................................................................................. xxii
CHAPTER 1. INTRODUCTION .................................................................................................... 1
CHAPTER 2. CHARACTERIZATION OF NOVEL SPLIT FLUORESCENT
PROTEINS AND QUANTITAVIVE ANALYSIS OF THEIR SELF-
ASSEMBLY PROCESS ................................................................................................... 56
vi

Fluorescence brightness of full-length FP and split-FP variants ............................ 70
Fluorescence photostability of full-length FPs and split-FPs ................................. 74
Folding and maturation rates of full-length FP variants .............................................. 77
Self-assembly process of split-FPs .............................................................................. 81
Live cells expression and imaging of split-FP variants ............................................... 92
Conclusion ................................................................................................................. 100
Material and Methods ................................................................................................ 102
Expression and purification of FPs ....................................................................... 102
Spectral acquisitions and photobleaching kinetics of FPs .................................... 102
Fluorescence lifetime measurements .................................................................... 103
Refolding and maturation kinetics of full-length FPs .......................................... 104
Self-assembly kinetic of split-FPs ........................................................................ 104
Anisotropy measurements .................................................................................... 105
Cell lines, cell labeling, and confocal imaging ..................................................... 106
Single molecule imaging and tracking ................................................................. 107
Immunoblot analysis ............................................................................................ 108
Introduction ............................................................................................................... 109
Guided clustering of metal NPs with controlled nanogaps using FP fragments ........ 112
GFP chromophore activation within plasmonic hot spots in metal NP clusters ........ 124
Site-directed assembly of AuNP clusters in live cells ............................................... 132
SERS microscopy imaging of metal NP clusters in targeted cells ............................ 139
Photoacoustic imaging of AuNP clusters on targeted cells ....................................... 144
Conclusion ................................................................................................................. 146
Material and Methods ................................................................................................ 148
Expression and purification of sGFP .................................................................... 148
Characterization of recombinant sGFP ................................................................ 149
Surface functionalization of nanoparticles with sGFP and M3 fragments ........... 150
Assessing the presence of sGFP on AuNPs ......................................................... 151
Dynamic light scattering and transmission electron microscopy of AuNPs ........ 152
Formation of nanoclusters by assembly of split-fluorescent protein
fragments and competition assay.......................................................................... 153
Statistical analyses of the size distribution of AuNP nanoclusters and
measurement of nanogap size between AuNPs .................................................... 153
Raman spectroscopy ............................................................................................. 155
Streptavidin titration on biotinylated M3-AuNPs and sGFP-AuNPs ................... 156
Fluorescence, dark field and total internal refection fluorescence
microscopy imaging of AuNPs targeted to avidin biomarkers in cells ................ 156
Scanning Electron Microscopy............................................................................. 157
SERS microscopy of AuNPs in cells .................................................................... 158
Photoacoustic microscopy imaging of AuNPs clusters on targeted cells ............. 159
Introduction ............................................................................................................... 161
SERS nanoprobes ...................................................................................................... 161
CHAPTER 3. TARGETED CELL IMAGING BY IN-SITU ASSEMBLY AND
ACTIVATION OF HOT SPOT SERS NANOPROBES USING SPLIT-
FLUORESCENT PROTEIN SCAFFOLDS ................................................................... 109
CHAPTER 4. FUTURE DIRECTIONS ...................................................................................... 161
vii

Detection sensitivity and selectivity .......................................................................... 162
Raman instrumentation and multimodality................................................................ 163
Summary and Conclusion .......................................................................................... 165
Instrumentation for Raman Spectroscopy ................................................................. 168
A1.1.1 Continuous Lasers ................................................................................................ 168
A1.1.2 Holographic notch filters ...................................................................................... 169
A1.1.3 Holographic gratings ............................................................................................ 170
A1.1.4 Array detectors ..................................................................................................... 170
A1.1.5 Renishaw inVia Raman system ............................................................................ 171
Surface Chemistry of Metal Nanostructures .............................................................. 172
A2.1.1 Surface modifications ........................................................................................... 172
A2.1.2 Poly(ethylene) Glycol (PEG) coating ................................................................... 172
A2.1.3 Hot spot formation of metal nanoparticles ........................................................... 173
SERS nanoprobes for biosensing............................................................................... 175
SERS nanoprobes for biomedical imaging ................................................................ 175
SERS nanoprobes for therapeutic and theranostic applications ................................ 177

.................................................................................................................................... 168
.................................................................................................................................... 172
.................................................................................................................................... 175
BIBLIOGRAPHY ........................................................................................................................ 182
viii

LIST OF FIGURES
Figure 1.1 Schematic of potential energy as a function of distance between atoms during
harmonic and anharmonic oscillation. ................................................................................ 4
Figure 1.2 Diagram of vibrational transitions of IR absorption/emission, and Raman and
Rayleigh scattering. ............................................................................................................ 7
Figure 1.3 Raman spectrum of coumarin molecule showing Rayleigh scattering at zero
position, both the Stokes and anti-Stokes Raman scattering at the positive and
negative sides of the spectrum [16]. ................................................................................... 8
Figure 1.4 Jablonski diagram illustrating the difference between IR, Raman scattering,
resonant Raman scattering, and fluorescence. .................................................................... 9
Figure 1.5 Illustration of the plasmon resonance difference between planar metal surface
and spherical nanoparticles. (a) Surface plasmon polariton and (b) localized
surface plasmon [22]. ........................................................................................................ 12
Figure 1.6 Dielectric function of bulk gold and silver materials. (a) Real (ε1) part of the
dielectric function gold (blue), and silver (red). (b) Imaginary (ε2) part of the
dielectric function of bulk gold (blue), and silver (red) [23]. ........................................... 12
Figure 1.7 Tunability of plasmonic behavior in different nanostructures. (a) Experimental
absorption spectra of Au nanospheres in different sizes. (b) Experimental
absorption spectra of Au nanorods with different aspect ratios (Adapted from
[41]. ................................................................................................................................... 14
Figure 1.8 Electric field distribution and SERS enhancement factor as a function of
interparticle distance. (a) Electric field distribution of single gold nanoparticle.
(b) Confined electric field at the hot spot due to the plasmon coupling of dimeric
nanoparticles with small separation. (c) SERS enhancement factor with respect to
the nanogap distance [44]. ................................................................................................ 16
Figure 1.9 Schematic of different functional groups on gold nanoparticle surface and
different functionalization methods [73]. ......................................................................... 21
Figure 1.10 Different conformations of oligonucleotides on gold nanoparticles based on
the surface density [84]. .................................................................................................... 23
Figure 1.11 Illustrations of possible SH-PEG conformations on gold nanoparticles. (a)
Number of PEG on 15 nm AuNPs exponentially decreases as a function of the
molecular weight (proportional with the length of PEG) of PEG, inset shows the
surface grafting density as a function of the molecular weight of PEG. (b) The
relationship between the number of PEG on AuNPs and the size of AuNP [67]. ............ 24
ix

Figure 1.12 SERS spectra of two different Raman reporters. (a) DTTC functionalized
gold nanoparticles are applied in vivo mouse model, and SERS signals are
detected from different locations in animal; subcutaneous injection spectrum
(green), deep tissue injection spectrum (blue), as controls pure tag spectrum (red)
and skin spectrum (black) [93]. (b) SERS spectra of 2-naphthalethiol on silver
nanoparticle embedded beads [96]. ................................................................................... 27
Figure 1.13 Surface coating strategies for SERS probes with (a) denatured BSA, (b) SH-
PEG, (c) amphiphilic diblock copolymer, (d) liposome and (e) silica shells [95]. ........... 28
Figure 1.14 Liposome encapsulated SERS nanoprobe. (a) Chemical structure of
malachite green isothiocyanate (MGITC). (b) TEM image of a bilayer-
encapsulated and MGITC functionalized gold nanoparticle, light gray ring
around the nanoparticle represents bilayer coating. (c) SERS spectrum of MGITC
on gold nanoparticles encapsulated with liposome [102]. ................................................ 29
Figure 1.15 SiO 2 encapsulation on PCEPE functionalized gold nanostructures. (a)
Transmission electron microcopy image of three gold nanoparticles
functionalized with the Raman reporter, PCEPE, and encapsulated with silica
shell (gray ring). (b) LSPR frequencies of L-shaped nanostructure measured by
dark-field Rayleigh scattering microscopy. (c) SERS spectrum of L-shaped
nanostructure. (d) Raman spectrum of PCEPE molecule. (e) Calculated Raman
spectrum of PCEPE and its chemical structure [109]. ...................................................... 30
Figure 1.16 DNA hybridization on gold nanoparticles. Thiol-conjugated oligonucleotides
functionalize gold nanoparticle surface. Upon introducing linking DNA duplex,
temperature dependent, reversible oligomerization creates controllable gold
nanoparticle aggregates [80]. ............................................................................................ 31
Figure 1.17 Synthesis and characterization of gold nanobridged nanogap particles (Au-
NNPs). (a) Scheme of forming gold layer on the template DNA-modified gold
nanoparticles. (b) TEM images of intermediates and fully synthesized Au-NNP.
(c) Comparison between uncontrollable nanoaggregates with non-uniform
nanogap size that cause non-uniform SERS signal (left) and controllable
nanostructures with well-defined nanogap that yield robust and quantitative
SERS signal (right) [110]. ................................................................................................ 32
Figure 1.18 Plasmonic biosensor design based on LSPR frequency, shape, and material.
(a) Gold nanoprisms modified with PEG and single stranded DNA (ssDNA)
exhibit extinction spectrum shift upon miR-X incubation in PBS buffer [124]. (b)
SEM image of vertical plasmonic nanoantenna (left), representative drawings of
cell/nanoantenna (center) and cell/planar SERS substrate (right) [125]. (c)
Schematic of reduced graphene oxide (rGO) separator between Au and Ag
nanoclusters on silicon (Si) substrate. Change in Raman spectra when normal
(202) and tumor (7402) cells were grown on rGO substrates; Raman spectra of
rGO on Si slide (green dash line, Ag@rGO@Au), normal live cells on rGO
substrate (blue solid line, 202@Ag@rGO@Au), and tumor cells on rGO
substrate (red dash line, 7402@Ag@rGO@Au) [126]. .................................................... 37
x

Figure 1.19 Single-walled carbon nanotube (SWNT) application on tumor bearing live
mice. White dotted line shows the vertical slice through the tumor. Ultrasound
(grey) and photoacoustic (green) images were overlapped at pre- and post-
injection time points. In order to observe the effect of the SWNTs, pre-injection
images were subtracted from post-injection images to create subtraction images.
The high photoacoustic signal (white arrows) in plain SWNT applied sample was
not observed in the subtraction image which indicates the presence of large blood
vessels. However, RGD peptide grafted SWNT applied sample shows high
photoacoustic signal in the subtraction image which shows the presence of
targeted SWNTs in the tumor region [128]. ..................................................................... 42
Figure 1.20 In vitro and in vivo photoacoustic imaging with gold nanoparticle as contrast
reagents. (a) Normalized photoacoustic signal intensity of 2D photoacoustic
images with respect to the RGD peptide surface density. Photoacoustic signal
intensity increases up to 1000 RGD peptide density, then the intensity decreases
with increased surface density. (b) Bright field and fluorescence microscopy
images of human prostate cancer cells (A-C). Three photoacoustic microscopy
(PAM) images show single cancer cells in red dashed circles (D-F). Scale bar:
100 μm, close-up scale bar: 25 μm [148]. ......................................................................... 43
Figure 1.21 Layout of optical systems. (a) 2D galvanometric mirror facilitated rapid laser
scanning system for Raman imaging [120]. (b) Hyperspectral Raman imaging
system with laser beam shaping module [150]. ................................................................ 45
Figure 1.22 Subcellular organelle targeting with gold nanobridged nanogap particles (Au-
NNPs). MB Raman reporter was used to modify Au-NNPs along with mPEG-
thiol and RGD peptide to target the cytoplasm; 44DP Raman reporter was used to
modify Au-NNPs along with mPEG-thiol, RGD and MLS to target mitochondria;
AB Raman reporter was used to functionalize the nucleus targeting Au-NNPs
with mPEG-thiol, RGD and NLS peptides. Representative Raman spectra from
inner cellular region at 3 h, 6 h, 12 h time points. Bright field and Raman images
were overlaid at 0 h, 3 h, 6 h, and 12 h time points [151]. ................................................ 46
Figure 1.23 The potential use of nanoshells/NIR treatment for photothermal ablation of
in vivo tumor tissues. (a) Calcein AM stained cells show cell viability after the
NIR laser treatment. However, nanoshells/NIR laser combination creates a clear
cell death. Fluorescein dextran staining is used to analyze the membrane
integrity. Nanoshell/NIR laser combination damages the cell membrane integrity
resulting in lack of cellular uptake of fluorescein dextran. However, only NIR
laser treatment does not cause enough damage on the membrane, thus due to the
cellular uptake fluorescence signal cannot be observed. (b) Histological analysis
on in vivo nanoshells/NIR laser treated tumor (transmissible venereal tumor
(TVT)) bearing mouse model. (1) Gross pathology after the nanoshells/NIR laser
treatment shows the tissue loss, (2) Silver staining reveals the localized
nanoshells, (3) Hematoxylin/eosin staining shows tissue damage in the similar
area nanoshells localized, (4) MRTI calculations show similar dimensions of
thermal damaged area [160]. ............................................................................................. 51
xi

Figure 1.24 Raman active molecule modified gold nanorods (NRs) for NIR detection and
photothermal therapy. (a) Bilateral human MDA-MB-435 tumor bearing nu/nu
mice were intratumorally injected with IR-792 modified gold NRs. (b) In vivo
Raman measurements reveals the Raman fingerprints of IR-792 modified gold
NRs. (c) Infrared thermographic map of mouse shows the significant temperature
elevation with 810 nm laser irradiation in the tumor region treated with IR-792-
coded gold NRs [164]. ...................................................................................................... 52
Figure 1.25 Core/Shell Pd/Au nanoplates as theranostic agents for in vivo photoacoustic
imaging and photothermal therapy. (a) In vivo photoacoustic imaging of Pd/Au
nanoplate injected 4T1 murine breast cancer bearing mice. Increased intensity of
photoacoustic signal is shown over time. (b) In vitro CT signal increases as a
function of increased concentration of Pd/Au nanoplates. (3) In vivo cross-
sectional CT image difference before and after injection of Pd/Au nanoplates and
(4) increased CT value after injection of Pd/Au nanoplates. (c) Infrared thermal
images of tumor bearing mice with Pd/Au treatment and control groups. Pd/Au
treated mice shows an increased temperature in tumor site upon 808 nm laser
irradiation. (d) Digital photographs of mice with only Pd/Au injection and Pd/Au
+ laser treatment [165]. ..................................................................................................... 53
Figure 2.1 Different application platforms of complementary split-GFP fragments. (a)
The schematic of the GFP (nano)polygon fabrication [167], (b) CALM imaging
of M3 peptide-coated coverslips before and after addition of split-GFP
fragments, and individual-complemented split-GFP molecules (red circles) [168],
(c) Identifying subcellular locations of synapses by GFP reconstitution across
synaptic partners (GRASP) [169]. .................................................................................... 57
Figure 2.2 Ribbon diagram of the enhanced green fluorescence protein (eGFP) crystal
structure (PDB entry: 2Y0G). The α-helices are shown in purple, the β-strands
are shown in yellow, the chromophore in the center α-helix is shown as ball-and-
stick model. ....................................................................................................................... 59
Figure 2.3 Illustration of chemical interactions between the GFP chromophore and the
surrounding water molecules and residues. (a) wtGFP and (b) S65T mutant [99]. .......... 61
Figure 2.4 Schematic of the interactions between the chromophore in CFP structure. The
chromophore is stabilized by van der Waals interactions (half circles) and
hydrogen bonds (dashed line). H148D mutations plays a role in planarity of the
chromophore [186]. .......................................................................................................... 64
Figure 2.5 Absorption and emission spectra of (a) split fluorescent proteins, and (b) full-
length fluorescent proteins. ............................................................................................... 66
Figure 2.6 Fluorescence lifetime (τ) comparison between full-length and split version of
the same fluorescent proteins. For YFP2, YFP3, CFP1 and CFP2 variants, only
the longest and generally dominant lifetime values are reported here. ○ represents
deprotonated B-state, ◊ represents neutral A-state and deprotonated B-state, *
represents neutral A-state and deprotonated I-state of the chromophore. ......................... 68
xii

Figure 2.7 Overlapped absorption spectra of the standard (A st) and the sample (A x). The
red circle shows the optimal excitation wavelength that has been used for
quantum yield calculations [201]. ..................................................................................... 71
Figure 2.8 Brightness comparison between all fluorescent proteins. ............................................ 72
Figure 2.9 Photobleaching kinetic of sGFP variants. (a) The model of irreversible
photobleaching and photoconvertible dark state reactions [205]. (b) Normalized
kinetic measurement for photobleaching and photoconvertible dark state
reactions of sGFPori (black), sGFP1 (red), sGFP2 (blue), and sGFP3 (green)
variants. (c) Comparison of faster photoconvertable dark state rate constants (left
panel), reverse photoconvertable dark state rate constants (center panel) and
slower irreversible photobleaching rate constants (right panel). ...................................... 77
Figure 2.10 Three-step mechanism of GFP chromophore formation model. Folding step
triggers the tripeptide chromophore cyclization which involves nucleophilic
attack of the amino group of Gly67 on the carbonyl group of Thr65. Cyclization
step is followed by an oxidation forms the mature GFP chromophore [172, 180]. .......... 77
Figure 2.11 Folding and maturation kinetic model of full-length FPs. Unfolded protein
with oxidized chromophore can be renatured to monitor folding kinetic
mechanism. Unfolded protein with reduced chromophore can be renatured and
reoxidized upon dilution from denaturant. (a) Folding (red) and maturation
(black) kinetics of flGFP2 were fitted by triexponential and single exponential
(green lines) growth functions, respectively. (b) Comparison of k fold1, k fold2, k mat,
and renaturation efficiencies of all the full-length FPs. .................................................... 81
Figure 2.12 Simple model of chromophore formation mechanism for split-FP fragments.
(a) The first condition of chromophore formation kinetic with sGFP2 titration
and 100 nM M3 peptide fragment. (b) The second condition of chromophore
formation kinetic with 0.5 μM sGFP2 and M3 peptide titration. ...................................... 82
Figure 2.13 Fluorescence anisotropy with respect to the concentration of ReAsH labeled
sGFP2. Endpoint fluorescence anisotropy was acquired from 0.5 μM, 1.25 μM,
2.5 μM, 5 μM, 10 μM, 20 μM ReAsH-sGFP2 complex. Increased concentration
of ReAsH-sGFP2 causes increased anisotropy values as an indication of forming
more dimeric structures. There is a concentration dependent monomer-dimer
equilibrium for in vitro sFP samples. ................................................................................ 84
Figure 2.14 Proposed model of split-FP chromophore formation mechanism under the
condition of sFP titration. (a) sGFP2 titration kinetics were fitted by
biexponential growth function, and pulled out k obs1 and k obs2 were plotted as a
function of calculated monomeric sGFP2 concentration based on the monomer-
dimer equilibrium ratio. k obs1 and k obs2 were assigned as k mat and k on, respectively.
(b) Comparison of k mat and k on for all split-FPs. NYD: Not yet determined..................... 87
Figure 2.15 Analysis of sFP chromophore formation kinetic based on the condition of M3
peptide titration. (a) Linear function fit on k obs1 rate constants from each kinetic
xiii

acquisition with different concentration of M3 peptide. The slope defines the k on
rate constant of binding step, and the y-intercept represents the k mat rate constant
of maturation step. (b) k obs2 and k obs3 rate constants from each kinetic acquisition
as a function of different M3 peptide concentrations. k obs2 and k obs3 rate constants
are assigned as k conf (conformational change of monomeric sGFP2) and k f
(forward reaction rate of monomer-dimer equilibrium), respectively. (c) The
fluorescence anisotropy kinetic measurements of sGFP2 only (red line), sGFP2
with excess 600 μM M3 peptide, and flGFP2 only (blue line). The single
exponential decay function was used to fit the anisotropy kinetic values (green
line). (d) Observed rate constants of each anisotropy kinetic measurement, k obsA,
from sGFP2 with different concentrations of M3 peptide. ............................................... 91
Figure 2.16 Potential model of split-FP chromophore formation mechanism. In the
condition of sFP titration (green dashed rectangle), the binding step is followed
by the maturation step. However, in the condition of M3 peptide titration, after
the binding step, the excess M3 peptide forces the reaction towards a yet not fully
defined pathway that requires a slow conformational change step on monomeric
structures before proceeding to maturation. The conformational change step is
followed by the maturation step. ....................................................................................... 92
Figure 2.17 The model for GPI anchored protein fusion with split-FP. ........................................ 93
Figure 2.18 Confocal imaging of GPI anchored split-FP fusion transfected U2OS cells.
The fluorescence images of the cells (top panel) and bright field images (bottom
panel) were shown. Scale bar: 15 μm. .............................................................................. 93
Figure 2.19 Immunoblot analysis of oligomerization of GPI anchored sGFP2 fusion on
the cell plasma membrane. ................................................................................................ 95
Figure 2.20 The model of complementation activated light microscopy (CALM) for the
complementation of sGFP-transmembrane fusion protein and M3 peptide
fragment on the cell plasma membrane [168]. .................................................................. 96
Figure 2.21 CALM imaging and diffusion trajectories of single molecule GPI anchored
split-GFP2 protein fusions. (a) During the addition of M3 peptide, individual
molecules in single frame upon complementation with M3 peptide fragment
(left), superresolved localizations of individual GPI anchor split-FP molecules
(center), diffusion trajectories of individual molecules. Scale bar: 10 μm. (b)
Brightness comparison of individual sGFPori and sGFP2 at the plasma
membrane. (c) Trajectory length comparison of individual sGFPori and sGFP2 at
the plasma membrane. (d-e) Time of first appearance of complemented sGFPori
and sGFP2 fragments upon introducing M3 peptide fragment. ........................................ 99
Figure 3.1 Characterization of AuNPs functionalized with split-fluorescent protein
fragments. (a) Schematic of surface modification on AuNPs by sGFP and M3
peptide fragments and formation of SERS active hot spot through self-assembly
and GFP complementation. (b) Comparison of the absorption spectra of bare
AuNPs (solid black), AuNPs coated with M3 peptides (M3-AuNPs, solid red),
xiv

and AuNPs coated with sGFP (sGFP-AuNPs, dash blue). Inset: Size exclusion
chromatography of M3-AuNPs. (c) Immuno-dot blot characterization of the
presence of full-length GFP or sGFP at the surface of AuNPs. (d) TEM images of
monodispersed AuNPs coated with M3 peptides or sGFP. Scale bar: 200 nm. (e)
DLS size distributions of bare AuNPs, M3-AuNPs and sGFP-AuNPs, coated
with different size PEGs. ................................................................................................ 113
Figure 3.2 Engineering sGFP for oriented binding on AuNPs. sGFP was expressed as a
recombinant protein with: (i) a 6xHis-tag for purification of the protein, (ii) a
GSS linker sequence, (iii) a thrombin cleavage site (LGLVPRC) to cut out the
6xHistag with thrombin after purification, (iv) a tetracysteine motif for oriented
binding of sGFP at the surface of AuNPs, (v) a flexible GGSGG linker domain to
limit conformation stiffness of the protein on AuNPs and (vi) the sGFP fragment
coding sequence. ............................................................................................................. 114
Figure 3.3 SDS-PAGE gel electrophoresis characterization of the expression and the
purification of the recombinant sGFP fragment. Lane 1: Molecular weight ladder.
Lane 2: Commercial sGFP (~25KDa). Lane 3: Unpurified cell lysate. Lane 4:
6xHistag purified sGFP (~26KDa). Lane 5: Thrombin cleaved sGFP. A few
higher molecular dimers of sGFP are observed (9%). .................................................... 114
Figure 3.4 Gel electrophoresis and ReAsh labeling to assess the presence and the activity
of the tetracysteine motif at the N-terminus of the sGFP fragment. Lane 1: sGFP.
Lane 2: ReAsh labeled sGFP. Lane 3: sGFP + M3 peptide fragment
complementation. Lane 4: ReAsh labeled sGFP + M3 peptide fragment
complementation. Notice that intramolecular FRET between complemented GFP
and ReAsh bound to the tetracysteine motif is observed in lane 4. ................................ 115
Figure 3.5 Agarose gel electrophoresis of bare, citrate-stabilized AuNPs (lane 1), sGFP-
AuNPs (lane 2), and full-length super-folder GFP-coated AuNPs (lane 3). The
direction of the electric field applied is indicated by the arrow. Notice that bare
AuNPs are unstable under these electrophoresis conditions and aggregate shortly
after entering the gel. ...................................................................................................... 116
Figure 3.6 TEM images of AuNPs functionalized with split-fluorescent protein variants
sYFP and sCFP. Scale bars: 200 nm. .............................................................................. 117
Figure 3.7 Dynamic light scattering characterization of AuNPs coated with sGFP or M3
peptide fragments and with different sizes of biotin-PEG moieties (5000 Da,
2000 Da, and 600 Da). .................................................................................................... 117
Figure 3.8 Formation of AuNP clusters by the assembly of FP fragments. (a) Agarose gel
electrophoresis of sGFP-AuNPs, M3-AuNPs and AuNP clusters and typical TEM
images of assembled clusters. White rectangles indicate the position of cluster
bands in gels. Scale bars: 20 nm. (b) TEM images of clusters formed by the
assembly of M3-AuNPs with sGFP-AuNPs, sYFP-AuNPs, or sCFP- AuNPs.
Scale bars: 200 nm. (c) Normalized AuNP cluster size distribution and fit by a
power law distribution model. (d) Size distribution of nanogaps formed by the
xv

assembly of FP fragments between AuNPs in clusters. The distribution is
Gaussian and centered at 2.1±0.5 nm. Inset: expected orientation of
complemented GFP at nanogaps. (e) TEM images of clusters formed by the
assembly of sGFP-gold nanorods with different sizes of M3-AuNPs. Scale bars:
20nm. .............................................................................................................................. 118
Figure 3.9 Agarose gel electrophoresis of 40 nm citrate-coated AuNPs functionalized
with sGFP and co-incubated with 10 nm oleic acid-coated AuNPs functionalized
with M3 peptides. TEM images of some of the nanoclusters observed after gel
extraction of the gel band corresponding to the nanoclusters (dash lines). Scale
bar: 20 nm. ...................................................................................................................... 119
Figure 3.10 Gel electrophoresis competition assay of the formation of AuNPs
nanoclusters using an excess of free and non-thiolated M3 peptide fragment.
Upon 12 hours co-incubation of 10 nm sGFP-AuNPs (lane 1) with 40 nm M3-
AuNPs (lane 2), a typical smeared band of AuNP nanoclusters is observed
together with the disappearance of the M3-AuNP band (lane 3). If a large excess
(100 µM) of free, non-cysteinilated M3 peptide fragment is added at the
beginning of the co-incubation, little to no smearing is observed and both sGFP-
AuNP and M3-AuNP bands remain intact (red arrow heads). This indicates that
the formation of AuNP clusters is solely driven by the self-assembly of split
fluorescent protein fragments appended at the surface of the nanoparticles. ................. 119
Figure 3.11 TEM images of compact nanoclusters formed after co-incubation of sGFP-
AuNPs, sYFP-AuNPs, or sCFP-AuNPs with M3-AuNPs. Scale bar: 200 nm. .............. 120
Figure 3.12 Kinetic analysis of the formation of AuNP clusters in solution by dynamic
light scattering (DLS). M3-AuNPs-PEG600 (40 nm) were co-incubated with
sGFP-AuNP-PEG600 (40 nm) in solution. The formation of AuNP clusters was
assessed by measuring the changes in the size distribution of the AuNPs using
DLS measurements at different time intervals (1, 2, 4, 6, 24 and 48 hours) and
comparing with their initial size distribution at time t=0 min. ........................................ 122
Figure 3.13 Statistical size distribution of unpurified AuNP clusters observed by TEM
and formed by co-incubating sGFP-AuNPs (40 nm) with M3-AuNPs (40 nm) for
12 hours (red bars). The cluster size distributions were determined over five
independent assembly experiments. 53% of all the AuNPs assemble into clusters
containing at least two AuNPs. The distribution is compared to the expected
Poisson distribution if the formation of the clusters was driven by random
interactions between AuNPs (black line, λ = 0.53). ........................................................ 123
Figure 3.14 Nanogap size measurement between two 40 nm AuNPs in a dimeric
nanocluster (left) and corresponding electron transmission intensity profile along
the blue line (right). The average of the maximum and minimum electron
transmission intensities is calculated from the intensity profile for each AuNP in
a cluster. The 50% value is used to estimate the gap edge for both AuNPs and to
measure the actual gap size. Scale bar: 20 nm. ............................................................... 124
xvi

Figure 3.15 TEM image of silver nanoclusters formed after co-incubation of 40 nm
sGFP-AgNPs and 40 nm M3-AgNPs. Scale bar: 200 nm. Inset: nanogap size
measurement of a dimeric AgNP cluster. Scale bar: 20 nm. .......................................... 124
Figure 3.16 Raman spectra of flGFP (310 μM), flYFP (232 μM) and flCFP (363 μM) in
aqueous buffer. For all three FP variants, the three main chromophore
fingerprints at 1530 cm
-1
, 1560 cm
-1
and 1660 cm
-1
are detected within ±10 cm
-1

of their expected spectral positions, together with weaker vibrational bands of the
chromophores and additional Raman bands attributed to the rest of the proteins.
Detected Raman bands were assigned in reference to previously corresponding
Raman shifts [112, 113, 254, 255]. λex: 785 nm, Pex: 3.33 mW/µm2, acquisition
time: 150 seconds. ........................................................................................................... 125
Figure 3.17 SERS spectra of full-length GFP, sGFP fragment, and complemented sGFP
on silver islands plasmonic substrates and colloidal AuNP clusters at different
pH. (a) SERS spectra of full-length GFP (blue), M3 peptide-complemented split-
GFP (red), and non-complemented sGFP fragment (green) on 5 nm silver island
substrates. (b) Liquid SERS spectra of assembled AuNP clusters at pH 8.0 (left)
and pH 6.0 (right). ........................................................................................................... 126
Figure 3.18 SERS spectra of M3-coated silver nanoparticles added on 5 nm silver island
plasmonic substrates shortly after co-incubation with the sGFP fragment. The
typical GFP chromophore fingerprints are observed at 1517 cm
-1
, 1555 cm
-1
and
1645 cm
-1
. *Assigned to the C-C-H in-plane deformation mode of the
chromophore’s phenol ring or the CH 2/CH 3 deformation mode of amino acids in
complemented sGFP. λ ex: 532 nm, P ex: 140 µW/µm
2
, acquisition time: 30 s. ................ 127
Figure 3.19 SERS spectra of individual and clustered NPs in buffer solutions. a) SERS
and Raman spectra of sGFP-AuNPs (top), M3-AuNPs (middle) and NaPT buffer
(bottom) at pH 8.0. In the buffer alone and in the absence of AuNP clusters, no
GFP chromophore vibrational signature is detected. λ ex: 785 nm, P ex: 20
mW/µm
2
, acquisition time: 30 s. b) Raman spectra of flGFP at pH 8.0 in TNG
buffer (top left) and at pH 6.0 in acetate buffer (top right) and corresponding
Raman spectra of pH 8.0 TNG buffer alone (bottom left) and pH 6.0 acetate
buffer alone (bottom right). The 1538 cm
-1
imidazolinone/exocyclic C=C mode
of the anionic GFP chromophore in flGFP is detected among vibrational modes
from the TNG buffer at pH 8.0. Change in pH induces the neutral form of the
chromophore and a shift of the 1538 cm
-1
mode toward 1549 cm
-1
. λ ex: 785 nm,
P ex: 20 mW/µm
2
, acquisition time: 30 s. c) SERS spectra of assembled 40 nm
M3-AgNP and sGFP-AgNP clusters at pH 8.0 (left) and pH 6.0 (center) and
Raman spectra of NaPT buffer at pH 6.0 (right) obtained for an excitation at 532
nm to optimize SERS responses. The anionic imidazolinone/exocyclic C=C
mode was detected at 1545 cm
-1
in the SERS spectrum of AgNP clusters at pH
8.0. The neutral mode was detected at 1570 cm
-1
in SERS spectra at pH 6.0. λ ex:
532 nm, P ex: 17 mW/µm
2
, acquisition time: 30 s. ........................................................... 129
Figure 3.20 Finite-difference time-domain calculated theoretical SERS enhancement
factors of the GFP imidazolinone/exocyclic C=C mode at 1530 cm
-1
for different
xvii

sizes of AuNP clusters with GFP-seeded nanogap dimensions of 2 or 4 nm. The
clusters are modeled as linear chain assemblies of 40 nm AuNPs with n = 2, 3, 4,
5 or 6 AuNPs per chain. .................................................................................................. 131
Figure 3.21 SERS spectra of sGFP-AuNPs, M3-AuNPs, and AuNP clusters assembled
for 12 hours from a 50 μl drop deposited on a SiO 2 wafer. The spectra were taken
after aggregations of the colloidal solutions following the evaporation created by
the laser. AmIII bands and CH 2/CH 3 scissor bending from sGFP and M3 peptides
were observed from all samples because of the artificial clustering. However,
GFP chromophore Raman fingerprints, which require binding and maturation
processes to form, were only observed from the complex sample. λ ex: 785 nm,
P ex: 33 mW/µm
2
, acquisition time: 60 s [112, 113, 254, 255]. ....................................... 132
Figure 3.22 Cell targeting and plasma membrane clustering of biotinylated AuNPs. (a)
Schematic of biotinylated M3-AuNPs and sGFP-AuNPs targeted to avidin fusion
proteins at the plasma membrane of cells and their assembly into SERS active
clusters. Not to scale. (b) Bright field (left), single frame TIRF microscopy image
(middle) and maximum intensity projection TIRF image from multiple frames
(∑I max, right) of biotinylated M3-AuNPs bound to GPI-avidin fusion proteins and
diffusing at the plasma membrane of live HeLa cells. Scale bar: 7 μm. c)
Scanning electron microscope images of U2OS cells co-expressing the
transmembrane and the GPI-anchored avidin fusion proteins and targeted with
biotinylated sGFP-AuNPs alone (top panel), biotinylated M3-AuNPs alone
(bottom panel) or both biotinylated sGFP-AuNPs and M3-AuNPs simultaneously
(middle panel). White arrows point towards endocytic membrane structures, blue
arrowheads point towards AuNPs monomers and red arrowheads point towards
some of the AuNP nanoclusters presented in insets. The plus and minus signs
identify the avidin-expressing and non-expressing cells, respectively. Scale bars:
2 μm (left panels), 200 nm (insets of left panels), 1 μm (right panels) and 100 nm
(insets of right panels). (d) Cluster size distributions of AuNPs on targeted U2OS
cells. ................................................................................................................................ 134
Figure 3.23 Reactivity of biotinylated AuNPs with streptavidin and specific targeting at
the plasma membrane of live cells expressing extracellular avidin fusion proteins.
a) 0.8% agarose gel shift assay after incubation of sGFP-AuNPs or M3-AuNPs
with decreasing concentrations of streptavidin (3.33 μM, 0.83 μM, 0.42 μM, 0.17
μM, 41.67 nM, 25 nM, 16.67 nM, 3.33 nM) for 45 min at room temperature.
While no shift is observed for non-biotinylated M3-AuNPs and sGFP-AuNPs,
discrete shifts are seen for the biotinylated NPs at high streptavidin
concentrations. This indicates that surface attached biotin-PEG moieties
effectively react with streptavidin despite the additional presence of M3 peptide
or sGFP fragments. b) Top: Fluorescence wide-field microscopy image of fixed
HeLa cells expressing the membrane avidin fusions and labeled with biotin-
Alexa594 fluorophore (yellow). Middle: Dark field microscopy image of the
cells (cyan). Bottom: Overlay of fluorescence and dark field images. Expressing
cell targeted with biotinylated sGFP-AuNPs display larger dark field signals than
non-targeted cells, which only show weak dark-field signals due to scattered light
from membranes and internal organelles. Scale bars 20μm. .......................................... 134
xviii

Figure 3.24 SEM images of U2OS cells expressing both avidin fusions and co-targeted
with AuNPs lacking surface biotin. a) Co-targeting of M3-AuNPs without biotin
together with sGFP-AuNP-biotin-PEG 600. b) Co-targeting of sGFP-AuNPs
without biotin together with M3-AuNP-biotin-PEG 600. While specific labeling of
expressing cells (+) is achieved, no plasma membrane clustering of nanoparticle
is observed if one of the two AuNPs lacks biotin-PEG 600. Blue arrowheads point
towards individual AuNPs presented in the insets. Scale bars: left panels: 2 μm,
right panels: 1 μm, insets: 100 nm. ................................................................................. 136
Figure 3.25 SEM images of U2OS cells expressing both avidin fusions and incubated
with sGFP-AuNP-biotin-PEG 2000 and M3-AuNP-biotin-PEG 2000 independently or
together. As with AuNPs modified with biotin-PEG 600 efficient clustering is
observed at the cell plasma membrane of expressing cells (+) for co-targeted
nanoparticles only. Blue and red arrowheads point, respectively, towards
individual or clustered AuNPs presented in the insets. Scale bars: Left panels: 2
μm, right panels:1 μm, insets: 100 nm. ........................................................................... 137
Figure 3.26 SEM images of HeLa cells expressing only the GPI-avidin fusion and
incubated with sGFP-AuNP-biotin-PEG 2000 and M3-AuNP-biotin-PEG 2000
independently (top) or together (bottom). Clusters are primary dimeric (red
arrowheads and insets) and residual monomeric AuNPs are observed at the
plasma membrane (white arrows). Scale bars: 1μm; insets: 100 nm. ............................. 138
Figure 3.27 Schematic of AuNP diffusion difference in NaPT buffer (top) and at the cell
plasma membrane (bottom). Cartesian coordinate axes represent both
translational and rotational diffusion in three-dimension inside the buffer system,
whereas it represents only translational diffusion in two-dimension at the cell
plasma membrane. .......................................................................................................... 139
Figure 3.28 Targeted SERS imaging of cells with split-FP assembled metal nanoclusters.
a) SERS microscopy images of fixed cells at the GFP chromophore 1527 cm
-1

imidazolinone/exocyclic C=C Raman mode and corresponding SERS spectra on
cells expressing the avidin biomarkers and targeted by biotinylated M3-AuNPs
and sGFP-AuNPs separately or simultaneously. Colored arrows in images point
toward single pixels whose SERS spectra are represented in matching colors. The
three typical GFP chromophore vibrational modes are indicated by dash lines on
spectra. b) SERS microscopy image of live cells co-targeted by biotinylated M3-
AuNPs and sGFP-AuNPs and reconstructed at a 1535±15 cm
-1
spectral window.
The deconvolved SERS spectrum corresponds to one pixel in the cell image as
indicated by the arrow. c) SERS image of live cells co-targeted by biotinylated
M3-AgNPs and sGFP-AgNPs and reconstructed at a 1550±15 cm
-1
spectral
window. The SERS spectrum corresponds to the individual pixel indicated by the
arrow in the cell image. d) SERS image of live cells in hypotonic buffer after co-
targeting of biotinylated M3-AuNPs and sGFP-AuNPs and reconstruction at a
1550±15 cm
-1
. The SERS spectrum corresponds to one pixel in the cell image as
indicated by the arrow. All scale bars: 10 µm. ............................................................... 142
Figure 3.29 Schematic of the laser-scanning photoacoustic microscopy system [148]. .............. 145
xix

Figure 3.30 . Photoacoustic imaging of in-situ assembled split-FP AuNP clusters on cells.
a) Photoacoustic microscopy images of individual U2OS cells among a 100%
confluent field after targeted clustering of biotinylated M3-AuNPs and sGFP-
AuNPs on cells that transiently express plasma membrane avidin biomarkers.
Scale bars: 50 µm (left) and 20 µm (right). b) Photoacoustic signal amplitudes
from similar fields of cells targeted with both M3-AuNPs and sGFP-AuNPs
(split-FP clustered) or with sGFP-AuNPs only (non-clustered). ***: p<0.01, T-
test. c) Photoacoustic signal amplitudes from targeted cells at increasing laser
excitation energy. Lines represent linear regression fit of the data. ................................ 145
Figure 4.1 Colloidal gold nanostars. (a) Vis/NIR spectra of gold nanostars synthesized by
gold nanoparticle (30nm) seeding. R is the ratio of gold salt concentration to gold
nanoparticle seed concentration. (b-d) TEM images of gold nanostars at R= 1.67,
11.25, and 20, respectively [278]. ................................................................................... 162
Figure 4.2 Illustration of general architecture of fiber-optics probes with bifurcation point
[279]. ............................................................................................................................... 164

Figure A1.1 Typical setup for Raman spectroscopy equipped with holographic
components, CCD detector, continuous wavelength laser. ............................................. 170
Figure A1.2 Renishaw inVia Raman microscope system. ........................................................... 171

Figure A2.1 PSPAA encapsulation on gold nanostructures functionalized with 2-
naphtalenethiol. TEM images of (a) monomeric, (b) dimeric, and (c) trimeric
gold nanostructures with PSPAA layer (gray ring). (d) UV/Vis spectra of PSPAA
encapsulated gold nanostructures, and differential centrifugation result to isolate
multimeric nanostructures groups [284]. ........................................................................ 173

Figure A3.1 Photoacousting imaging and PET scans on tumor that are treated with Au-
tripods. (a) Photoacoustic images (PAI) of targeted integrins on tumor cell
membrane pre- and post-injections of RGD-Au-tripods at 1h, 2h, 4h time points
(left panel). After blocking the integrins on tumor cell membrane, photoacoustic
images of pre- and post-injections of RGD-Au-tripods. (b) Small animal PET
images of integrin positive tumor bearing mice injected with RGD-Au-tripods to
target integrins [289]. ...................................................................................................... 177
Figure A3.2 Nanoshells/NIR laser combination with dark field imaging on breast
adenocarcinoma cells expressing HER-2. Scattered-based dark field imaging to
assess HER2 expression (top row), calcein staining to analyze cell viability after
laser treatment (center row), silver staining to localize nanoshells bindings
(bottom row). In dark field imaging, anti-HER2 modified nanoshells is observed,
this indicates cells with HER2 expression can be targeted with nanoshells (top
row, right column). After NIR laser treatment, cytotoxicity (dark spot) is
observed only from nanoshells treated sample (center row, right column). Silver
staining reveals the presence nanoshells on the cells (bottom row right column)
[290]. ............................................................................................................................... 178
xx

Figure A3.3 Tumor tissue response to photothermal therapy monitored by
bioluminescence imaging. Luciferase transfected tumor cells create
bioluminescence signal. Representative mice of each condition for certain days
of the evaluation. Nanomatryoshka (NM) injected mice do not show tumor tissue
recurrence even 60 days after the laser treatment. However, nanoshells (NS)
injected mice show tumor recurrence 2 weeks after the laser treatment. [121]. ............. 179
Figure A3.4 In vitro and in vivo characterization of FA-NT-AuNBPs treated MCF-7 cells
and MCF-7 tumor bearing mice. (a) Fluorescence images of MCF-7 cells (A)
without and (B) with FA-NT-AuBPs treatment under 808 nm laser irradiation at
different time points. Calcein AM/ethidium homodimer-1 staining is used to
differentiate live/dead cells. Calcein AM with green fluorescence indicates live,
ethidium homodimer-1 with red fluorescence indicates dead cells. There is a
clear photothermal ablation with FA-NT-AuBPs treatment upon 808 nm laser
irradiation. (b) IR thermal imaging of MCF-7 tumor bearing mice with (A) PBS
control injection and (B) FA-NT-AuBPs injection under 808 nm laser irradiation
at 0 min, 1 min, 3 min, 5 min. Fa-NT-AuBPs injected mice group shows a
significant temperature increase in the tumor region upon laser irradiation. (c)
Digital photographs of MCF-7 tumor bearing mice at 0 day and 14 days time
points from four different groups; a-PBS, b-PBS+NIR, c-FA-NT-AuBPs, d-FA-
NT-AuBPs+NIR [291]. ................................................................................................... 180

xxi

LIST OF TABLES
Table 2.1 Fluorescence properties of full-length and split fluorescent proteins with
mutations they carry. ......................................................................................................... 64
Table 3.1 Frequency of cluster size distribution for sGFP-AuNPs and M3-AuNPs
functionalized with biotin-PEG 600 or non-biotinylated-PEG 600 and simultaneously
co-targeted or separately targeted on U2OS cells. Cells co-express both the GPI-
and transmembrane avidin fusions. ................................................................................. 135

xxii

ABSTRACT
In this thesis, a variety of novel split-fluorescent proteins (split-FPs) were developed and
employed to design new surface enhanced Raman scattering (SERS) imaging nanoprobes capable
of self-assembling into photonically active hot spots directly in live cells. These original optical
probes use complementary split-FP fragments as molecular glue to control the assembly of gold
or silver metal nanoparticles (AuNP or AgNP) clusters, whose tunable electronic and plasmonic
properties have gained increasing interests for cellular and in vivo imaging applications. By
appending complementary split-FP fragments at the surface of AuNPs or AgNPs, full-length and
mature FPs can be reconstructed at the interface between clustered NPs and uniform plasmonic
hot spots, a few nanometer in size, are generated. These hot spots provide massive near-field
electromagnetic enhancements of the FP chromophore peculiar Raman fingerprints, allowing
highly specific and sensitive SERS imaging of targeted cells. Because both complementary split-
FP fragments are required to induce chromophore cyclization and activation of the FP Raman
fingerprints, the detection selectivity of targeted cells by SERS imaging is significantly increased.
Furthermore, FP-driven in situ metal nanoassemblies yield strong photoacoustic signal, allowing
the FP/NP hybrid nanoclusters to serve as promising contrast agents for multimodal SERS and
photoacoustic microscopy imaging with single cell sensitivity.
A comprehensive background on the different components of this thesis work is provided
in the introductory CHAPTER 1 together with more detailed explanations available in Annex
sections. In the first part of CHAPTER 1, physical models of Raman spectroscopy and
fundamentals of SERS are discussed with an emphasis on the advantages of SERS techniques. In
the second part, we focus on the surface chemistry of metal nanoparticles, on the surface
functional groups related to this project, and on the Raman reporters and their prominence for hot
spot formation. To finish CHAPTER 1, we discuss the use of SERS nanoprobes as contrast
xxiii

agents for biomedical imaging and therapeutic/theranostic applications. CHAPTER 2 describes
the development and the thorough characterization of novel split-FPs that we have used as Raman
reporters in this thesis work. A quantitative analysis of the split-FP self-assembly process is
presented through a comparative study of multiple spectral variants. The details of this self-
assembly process hold great importance for this project because in addition to being Raman
reporters upon complementation, split-FP fragments serve as complementary scaffolds to drive
the formation of nanoassemblies. In CHAPTER 3, the surface functionalization of metal
nanoparticles with complementary split-FP fragments and their assembly on targeted cells in vitro
are described. A comprehensive characterization of the FP/NP hybrid probes by transmission
electron microscopy analysis, dynamic light scattering technique, gel electrophoresis, and Raman
spectroscopy is presented. Furthermore, cellular application of the SERS probes is studied by
scanning electron microscopy analysis, SERS imaging and photoacoustic microscopy. Finally, in
0, future directions are proposed, such as the influence of nanomaterial geometry and the
potential alterations of surface functional groups for dual targeting of different cell plasma
membrane receptors. In this final chapter, prospective ideas to integrate different imaging
techniques to improve in vivo experiment are also presented.

1

CHAPTER 1. INTRODUCTION

Fundamentals of Raman Spectroscopy
Introduction
Raman spectroscopy is an optical characterization tool, which provides insights into the
structural identity of a molecule based on its vibrational fingerprints. In 1928, Raman and
Krishnan demonstrated the frequency-shifted scattered light effect on 60 different liquids where
the scattering is independent from the incident light, hence the Raman effect is distinguished from
fluorescence [1]. For many decades, Raman spectroscopy was an inefficient technique and
limited on in situ applications due to the lack of technical innovations. Since Raman scattering
depends on a frequency shift from the incident light frequency, it is essential to have a narrow,
monochromatic and coherent light source. After the advent of the laser in 1960, Raman
spectroscopy started to grow and became available in many laboratories. Following
improvements on various instrumentations rendered the simplification of the Raman systems
including the use of notch filters, holographic gratings, and the availability of cooled charge
coupled device (CCD) detectors for enhanced signal to noise ratio and high quantum efficiency
(counts/photon) [2]. Raman spectroscopy became significantly simpler to implement.
Furthermore, the use of interferometers and near-infrared (near-IR) lasers reduced the
fluorescence interference and thermal damage on the samples [3]. Until these technical
improvements are in place, Raman spectroscopy was considered as a less sensitive and more
time-consuming technique compared to the IR spectroscopy which has been revolutionized as
Fourier transform infrared (FT-IR) spectrometry afterwards. Raman cross section is in between
10
-31
-10
-29
cm
2
/molecule, on the contrary fluorescence cross section can reach 10
-16
cm
2
/molecule
[4]. Therefore, the major drawbacks of Raman spectroscopy are the probability of observing one
2

Raman scattered photon is ~10
-6
and the fluorescence from the sample often interferes with high
signal to noise detections. Essentially, Raman and IR spectrometry are complementary ways of
measuring quantized vibrational energy transitions of molecules; the former measures the
scattered radiation of a molecule at relative frequencies while the latter measures the absorbed
radiation of a molecule at absolute frequencies. The vibrations of a molecule might be IR active,
if the dipole moment of the molecule changes during vibrations, or it might be Raman active, if
the polarizability of the molecule changes during vibrations. Moreover, homonuclear diatomic
molecules do not have a dipole moment, so the vibration is IR inactive. On the other hand, both
homonuclear and heteronuclear molecules have a change in polarizability, therefore the vibrations
of these molecules are both Raman active. Although, in polyatomic molecules, due to the
increased complexity of vibrations, detecting the alterations in dipole moment and polarizability
of the molecules becomes complicated. For instance, asymmetric stretches of a linear triatomic
molecule are IR active but Raman inactive. Whereas the second derivatives of these asymmetric
vibrations become Raman active and rarely visible in nonresonant Raman scattering since these
bands are significantly weaker. To ameliorate the problem, the resonant Raman (RR) scattering
comes into play where the frequency of the incident light is the same as the electronic absorption
wavelength of the molecule, and this increases the Raman cross section. In resonant Raman
spectroscopy (RRS), Raman spectra are dominated by the symmetric vibrations of the molecule,
also the symmetric vibrations from the second derivative of the asymmetric vibrations can be
excited. While the RRS was drawing a lot of interests due to the innovations on tunable lasers,
significant improvements in Raman detection has emerged by the use of metal surfaces to
enhance the Raman signals. Fleischmann et al. has recorded the first experimental results of
surface enhanced Raman spectroscopy (SERS) on pyridine adsorbed silver electrodes [5]. Two
effects have been described to explain the enhanced Raman signals of molecules on metal
3

surfaces. The first one is the increased molecular Raman scattering cross section due to the
electric field enhancement by Jeanmaire and Van Duyne [6] and the second one is the coupling of
metal surface plasmon polariton with the adsorbed molecules [7]. The combination of resonant
Raman scattering and SERS results in extreme sensitivity and allows to work on low
concentration samples, namely surface enhanced resonant Raman spectroscopy (SERRS) [8].
Early studies have used metal electrodes etched with chemicals resulting in roughened metal
surface which makes the excitation of surface plasmon modes possible [5, 9]. Later on, colloidal
silver and gold particles have been used for SERS studies and this confirmed that the plasma
oscillations were because of the uneven structures on the electrochemically roughened metal
electrodes [10] and this study leads to a wide variety of plasmonic materials for the use of SERS
experiments.
In this chapter, we give detailed explanation of the physical model of Raman
spectroscopy and the instrumentation, followed by fundamentals of surface enhanced Raman
spectroscopy, description of the localized surface plasmon resonance, plasmonic coupling, and
different plasmonic materials.

Physical Model of Raman Spectroscopy
Transitions between vibrational energy levels of a molecule give rise to IR and Raman
spectra. A molecule with N number of atoms has 3N degrees of freedom; three for translational
motion (x-, y-, and z-axes), three for rotational motion (x-, y-, and z-axes), and the rest gives the
vibrational modes (different ways of atomic vibrations in a molecule) of the molecule. In each
vibrational mode (i), all the atoms oscillate at a certain frequency (ν i) and creates a harmonic
displacement. In Figure 1.1, the potential energy, V(r), as a function of the interatomic distance (r)
4

is shown. The simple harmonic oscillation can be explained by Hooke’s Law, where the
vibrational energy states, V iν, is equal to:
𝑉 𝑖𝜈
= ℎ𝜈 𝑖 (𝜐 𝑖 +
1
2
)
1.1

In Equation 1.1, h is Planck’s constant, ν i is the vibrational mode frequency, and υ i is the
vibrational quantum number of ith mode. However, as it is shown in Equation 1.2, simple
harmonic oscillator model applies only to the small vibrational quantum numbers. In practice,
when the υ i increases, the anharmonic (Morse-type) model can describe the potential energy of
the vibrational states, V iν:

𝑉 𝑖𝜈
= ℎ𝜈 𝑖 (𝜐 𝑖 +
1
2
) + ℎ𝜈 𝑖 𝑥 𝑖 (𝜐 𝑖 +
1
2
)
2

1.2

where x i is the anharmonicity constant. Unlike the simple harmonic oscillation, anharmonic
oscillations include overtones and combinations as well as fundamental transitions.

Figure 1.1 Schematic of potential energy as a function of distance between atoms during harmonic and anharmonic
oscillation.

5

A few vibrational modes of a molecule show large atomic displacement and the rest of
the modes are often stationary. The large displacements usually belong to a certain functional
group which can be used as an indicative of the molecule through specific spectral signatures.
In a molecule, when the centers of positive and negative charges do not coincide, this
creates an electric dipole which is an efficient source of electromagnetic radiation at a frequency
of ν M. This oscillating dipole also absorbs at the same frequency which leads to excitation of the
molecular vibrations. The oscillation of the induced electric dipole is also related with the
scattering of light. When an external electric field (incident radiation) is applied to a molecule, it
polarizes the molecule by distortion of the electron cloud around the positive nuclei in opposite
directions and creates a dipole moment. The isotropic polarizability of the molecules can be
explained by Equation 1.3 which defines a scalar system where the applied electric field induces
parallel polarization to the field.
𝜇 ′ = 𝛼𝐸 1.3

In Equation 1.3, μ ʹ is induced dipole moment vector, E is the electric field vector, α rank two
polarizability.
Whereas, the anisotropic polarizability can be described as a rank two tensor with a 3x3
matrix, Equation 1.4. Although isotropic polarizability cannot cause a frequency-shifted light
scattering, anisotropic polarizability creates a frequency-shifted light and gives rise to Raman
scattering.
[
𝜇 𝑥 ′
𝜇 𝑦 ′
𝜇 𝑧 ′
] = [
𝛼 𝑥𝑥
𝛼 𝑥𝑦
𝛼 𝑥𝑧
𝛼 𝑦𝑥
𝛼 𝑦𝑦
𝛼 𝑦𝑧
𝛼 𝑧𝑥
𝛼 𝑧𝑦
𝛼 𝑧𝑧
] [
𝐸 𝑥 𝐸 𝑦 𝐸 𝑧 ]
1.4

If all the polarizability tensor components are equal to zero, the vibrations are Raman inactive; if
only one component is nonzero, those vibrations are Raman active.
6

When a sample is illuminated by a beam of monochromatic light at ν 0 frequency, a small
portion of the incident light is scattered. Scattering refers to deflection of the light from the path
of the incident light. If the scattered light is observed at the same frequency as that of the incident
light, the scattering is described as elastic scattering (Rayleigh scattering). If it is observed at a
different frequency than the incident light, the scattering is described as inelastic scattering [11].
In inelastic scattering, the new frequencies (νʹ) appear as pairs in the spectrum that are shifted
with respect to the Rayleigh scattering (ν 0) with νʹ = ν 0 ± ν m, where ν m represents the frequency
difference between the vibrational states. The appearance of new frequencies in scattered light is
called Raman scattering, and the frequencies are called Raman bands or Raman shifts. If the
energy of a scattered photon (hν 0-hν m) is less than the energy of an incident photon (hν 0), this
Raman band is referred as Stokes-Raman shift. If an electron already at excited vibrational state
interacts with an incident photon (hν 0) and creates a higher energy scattered photon (hν 0+hν m),
this Raman band is called as anti-Stokes Raman shift (Figure 1.2). In IR absorption and emission,
a simple one-photon process occurs where either one photon is absorbed or emitted during a
direct transition between vibrational energy levels. On the other hand, Rayleigh and Raman
scattering involve two transitions between the vibrational energy levels and virtual states where
one photon is annihilated from the incident radiation and another photon is created either at the
same energy (Rayleigh scattering), at the higher energy (anti-stokes Raman), or at the lower
energy (Stokes Raman). These virtual states have picosecond lifetime and they are not stationary
eigenstates where the system remains the same as a function of time [12].
7

Figure 1.2 Diagram of vibrational transitions of IR absorption/emission, and Raman and Rayleigh scattering.

The molecular vibration frequencies are represented by the frequency shifts (Δν s) in the
Stokes and anti-Stokes Raman spectra (Figure 1.3). Raman shifts are in the range from 0 to 4000
cm
-1
which also covers IR and far-infrared (FIR) spectral regions. Compared to anti-Stokes
Raman shifts, Stokes Raman shifts exhibit higher intensity in experiments where off-resonance
conditions are applied at room temperature. However, when the anti-Stokes Raman shifts are
normalized by the Boltzmann constant and the temperature, the ratio between the Stokes and anti-
Stokes Raman shifts becomes 1 [13]. Hence, the ratio between Stokes and anti-Stokes Raman
intensities can be used to determine the sample temperature;

𝐼 𝑆 𝐼 𝐴𝑆
= 𝑒𝑥𝑝 (
ℎ𝜈 𝑘 𝐵 𝑇 )

1.5

8

I S is Stokes-Raman intensity, I AS is anti-Stokes Raman intensity, h is the Planck’s constant
(6.62608 x 10
-34
J s), ν is the difference of photon frequency, k B is the Boltzmann constant
(1.38064 x 10
-23
m
2
kg s
2
K
-1
), T is the temperature of the sample [14].

Resonance Raman Spectroscopy and Surface Enhanced Raman Spectroscopy
Raman scattering is independent from excitation laser wavelengths, however at certain
excitation laser frequencies, the Raman band intensities are strongly modulated. Even though
Raman emitted photons usually have low intensity, they can be enhanced by several different
methods. One way is using an excitation wavelength in the range of electronic absorption band of
the molecule. This technique is called resonance Raman scattering (RRS) and first experimental
observation was achieved by Harrand and Lennuier on dichloronitrobenzen where strong Raman
signals have been observed when the excitation wavelength was in the absorption band [15].

Figure 1.3 Raman spectrum of coumarin molecule showing Rayleigh scattering at zero position, both the Stokes and
anti-Stokes Raman scattering at the positive and negative sides of the spectrum [16].

9

Furthermore, they reported that corresponding functional groups to the enhanced Raman bands
are the ones responsible from the absorption band. The rest of the vibrational bands remain
relatively weak due to off-resonance conditions. The resonance Raman enhancement of the
certain vibrational modes gives an opportunity to work with diluted samples. Hence, RRS
provides more sensitive and selective spectral detection compared to the basic Raman
spectroscopy. In resonant Raman spectroscopy, the electron on the ground state can be excited to
a vibrational quantum number of a higher eigenstate (Figure 1.4). However, as seen in Figure 1.4,
when a molecule or a chromophore absorbs an incident radiation, fluorescence might occur which
creates a background and mostly dominates the Raman signals.

Figure 1.4 Jablonski diagram illustrating the difference between IR, Raman scattering, resonant Raman scattering, and
fluorescence.

When an electric field or a laser beam interacts with a metal surface, the reflected beam
of light from the metal surface creates an electric standing wave field with a high amplitude
which has been used to observe Raman bands of benzoic acid on flat silver surface with 10 fold
10

or less enhancement on the signal intensity [17]. On the other hand, if the Raman active molecule
is adsorbed on a roughened metal surface, the incident laser radiation creates a confined and
enhanced electromagnetic (EM) field close to the surface. The enhanced EM field enhances the
Raman signal due to the proportional relationship between the Raman intensity and the square of
the incident field strength. This phenomenon is called surface enhanced Raman scattering (SERS)
which is another way to enhance the Raman signals. It has been shown that the roughened silver
surface has created factor of 10
6
enhancement compared to the same concentration of pyridine in
the absence of the surface [6, 18]. The interaction between the adsorbate and the surface plasmon
polaritons gives rise to the SERS phenomena [7]. On smooth metal surfaces, surface plasmons
(delocalized charge oscillation) radiate only in parallel direction to the surface. However, on
roughened metal surfaces, surface plasmons can move in both parallel and perpendicular
directions. These surface plasmons can be excited by an incident photon and this results in
enhanced electromagnetic field due to the coherent oscillations of conduction electrons. When the
enhanced electromagnetic field interacts with the adsorbate, it enhances the Raman scattering.
The electromagnetic field enhancement is the major component of the enhancement mechanism.
Furthermore, it has been shown that the charge transfer mechanism also contributes to the
enhancement factor within the first monolayer of the adsorbate, even though it is hard to
distinguish these two mechanisms [19]. This chemical enhancement requires a physisorption or
chemisorption of the molecule to the metal surface.

Plasmonic materials
SERS efficiency highly depends on the properties of the plasmonic surfaces. Metals
interact with the incident light differently in bulk sizes compared to their nanoscale sizes. When
the incident photon frequency and the surface plasmon resonance frequency are equal, the metal
11

absorbs the light and form surface plasmon polaritons at the metal-dielectric interface. If the
frequency of the incident photon is lower than the plasmon resonance, the metal surface reflects
the light, that’s why most of the metals are shiny under visible light. If the frequency of the
incident photon is higher than the plasmon resonance, the metal transmits the light. When the size
of the metal is smaller than the wavelength of the incident light, the conduction electrons of the
metal follow the electric field of the incident light which results in induced dipole moment in the
metal particle. This effect is called surface plasmon resonance (SPR) for the planar metal
surfaces, and localize surface plasmon resonance (LSPR) for different geometrical metal
nanoparticles (Figure 1.5). However, highly mobile electrons in a metal cannot follow the
incident light at high frequencies in the visible range. Therefore, incident light frequency and the
electron mobility play an important role on the optical and reflective properties of the metal [20].
Dielectric function (dielectric permittivity) of the materials provides information about their
optical response. In complex dielectric function (ε(ω) = ε 1 + i ε 2), the real part (ε 1) describes the
reflection (electric flux concentration in the material), and imaginary part (i ε 2) describes the
absorption and optical loss in the system which corresponds to LSPR (or SPR) [21].

12

Figure 1.5 Illustration of the plasmon resonance difference between planar metal surface and spherical nanoparticles.
(a) Surface plasmon polariton and (b) localized surface plasmon [22].

Figure 1.6 Dielectric function of bulk gold and silver materials. (a) Real (ε1) part of the dielectric function gold (blue),
and silver (red). (b) Imaginary (ε2) part of the dielectric function of bulk gold (blue), and silver (red) [23].

The dielectric function comparison of gold and silver materials in bulk form can be seen
in Figure 1.6 [23]. Both materials exhibit a decrease in the real part of the dielectric function (ε 1)
13

at the lower frequency than the plasmon frequency which indicates lower electron flux through
the material and higher reflection characteristic. Ag shows lower imaginary dielectric function
(ε 2) across 300-1100 nm range than Au which implies less optical loss (plasmon damping) and
higher scattering efficiency.
Gold (Au), silver (Ag), and copper (Cu) materials are the most commonly used SERS
substrates because the LSPRs of these materials are in the visible and near infrared region where
most of the Raman experiments are conducted [24]. Highly enhanced pyridine Raman signal was
acquired on electrochemically etched silver electrode [5]. Compared to other materials, the largest
surface enhancement was obtained from silver electrodes [25]. The degree of roughness on the
electrodes is an important parameter on the Raman band intensities and it can be controlled by the
electrochemical cycles [26]. It has been shown that when the negative potential is increased, the
Raman intensity of the adsorbed molecule is decreased since oxidized silver ions becomes
soluble. Whereas, when the potential is low, the relative intensity of Raman band is higher due to
reduced silver ions [27]. Moreover, the applied potential on the electrode during the Raman
acquisition can affect the adsorption and the orientation of the molecule which changes the
selection rules [28, 29]. Increased potential can also affect the surface coverage towards removing
the adsorbate and silver from the surface [29]. As a consequence, reproducibility is a long-
standing problem for electrode base SERS applications. On the contrary, colloidal nanoparticles
as a SERS substrate became more attractive due to relatively inexpensive and reproducible
productions of nanoparticles [10, 30-32]. The first demonstration of utilizing colloidal
nanoparticles for SERS was provided by Creighton et al [10] and involved pyridine adsorbed on
Au and Ag nanoparticles. Further improvement on SERS signal amplification, ~10
10
, due to the
highly confined electromagnetic field was achieved by exotic shaped nanostructures with sharp
edges and corners [33, 34]. Various number of geometrical nanostructures with different
14

materials in different sizes can be created to tune their plasmonic behavior (Figure 1.7) [35-37].
Even though nanoparticle surface modification to increase the electrostatic/electrochemical
attraction between nanoparticles is challenging, controlled surface functionalization strategies of
colloidal nanoparticles have been developed and they have allowed bottom-up formation of three-
dimensional structures for different application purposes, such as gold-DNA origami
nanostructures [38] and 3D plasmon rulers [39]. Recently, mesoscale superlattice structures were
created using nanoscale Au particles which are precisely assembled in three-dimensions by DNA
oligomerization [40].

Figure 1.7 Tunability of plasmonic behavior in different nanostructures. (a) Experimental absorption spectra of Au
nanospheres in different sizes. (b) Experimental absorption spectra of Au nanorods with different aspect ratios
(Adapted from [41].

Controlling the interparticle distance provides an advanced SERS enhancement through
plasmon coupling of the individual nanoparticles [42]. In a single nanoparticle, the enhanced
electric field is confined within a nanometer range near the surface (Figure 1.8 a), and it decays
dramatically thereafter [43, 44]. The enhanced electric field can be further improved by bringing
two nanoparticles into close proximity. This leads to an interaction between surface plasmon
15

resonances of two nanoparticles where the overall plasmon frequency shows red-shift compared
to the single nanoparticle [45]. The near field coupling creates a strong electric field at the
interparticle junction (Figure 1.8 b) and these nanogaps with enhanced electric field are called
“hot spots” [44, 46]. The strength of plasmonic coupling is proportional to the SERS
enhancement factor, but inversely proportional to the interparticle distance (Figure 1.8 c) [44, 47,
48]. As the nanogap distance decreases, the individual plasmons within the nanoparticles
hybridize and this creates dimer plasmons with finite dipole moments. On the other hand, when
the nanoparticle separation increases, the plasmon interaction between nanoparticles becomes
weak and plasmon energy resembles to the single nanoparticle [48, 49]. In the plasmon
hybridization model, the polarization of the incident light has a significant role on obtaining
highly enhanced and confined electric field. When the polarization is parallel to the axis of
dimeric interparticle axis, the plasmons couple strongly and plasmon frequency shows a red shift.
Whereas, if the incident light is polarized perpendicularly to the dimeric interparticle axis, the
plasmon coupling response becomes insensitive and the frequency shows a small blue shift [46,
50]. In this thesis work, it is essential to control the nanoparticle gap size that yields strongly
enhanced electromagnetic field to obtain high signal to noise detection in biological
environments.
The details about the instrumentation of Raman spectroscopy can be found in ANNEX 1.
16

Figure 1.8 Electric field distribution and SERS enhancement factor as a function of interparticle distance. (a) Electric
field distribution of single gold nanoparticle. (b) Confined electric field at the hot spot due to the plasmon coupling of
dimeric nanoparticles with small separation. (c) SERS enhancement factor with respect to the nanogap distance [44].

Conclusion
In this section, the development of Raman spectroscopy and how it has become a
common analytical tool was summarized and the physical model of this technique was described.
The differences between Rayleigh scattering, Stokes and anti-Stokes Raman scatterings, resonant
Raman scattering (RRS) and surface enhanced Raman scattering (SERS) were explained.
Furthermore, the importance of plasmonic materials on SERS applications were described and the
formation of surface plasmon resonance (SPR) and localized surface plasmon resonance (LSPR)
of plasmonic materials based on their dielectric functions was delineated. We illustrated how
different physical manipulations on plasmonic nanostructures can alter their optical and
plasmonic behaviors. In ANNEX 1, the instrument setup was outlined and the important optical
components were defined to increase the efficiency of Raman systems.
17

The fundamentals of Raman spectroscopy and plasmonic behaviors of metallic materials
covered in this chapter is essential to understand the Raman imaging technique which was used
for targeted cells in this thesis. Hot spot formation yields high SERS enhancement which is
extremely critical to overcome the fluorescence background in biological samples. High signal to
noise detection can also be improved by using NIR wavelengths which reduce the fluorescence
background as well as the damage on the specimen due to the lower energy photons. In this
thesis, split fluorescence protein fragments have been used for surface modification of plasmonic
metal nanoparticles to form hot spots. In the following section, the surface chemistry of
nanoparticles will be explained in detail to create multimeric nanostructures resulting in confined
electromagnetic field enhancement for SERS applications.

18

Surface Chemistry of Metal Nanoparticles
1.2.1 Introduction
As discussed in the previous section, the use of metal nanoparticles provides a simple
means to significantly enhance the Raman signals of molecules adsorbed on their surfaces.
Therefore, there is a growing interest in employing metal nanoparticles for Raman detection by
SERS because electronic and optical properties of metal nanoparticles can be optimized for
Raman spectroscopy by simply tuning their size, shape, composition, and surface coating. The
ease on controlling electronic and optical properties makes nanoparticles attractive units for
developing various applications, such as biosensors, medical diagnostics and therapeutics,
photonic circuits, nanoscale energy and electronic devices [51-56]. There are a wide variety of
protocols about synthetic production of nanoparticles within a narrow size distribution [57, 58].
When the nanoparticles are synthesized by “wet chemistry”, they require a surface
functionalization to create a protection layer against surrounding medium. This protection layer
acts as a temporary coating to stabilize and prevent the aggregation of nanoparticles by making
them resistant to external factors, for instance nonspecific adsorption of surrounding molecules,
or buffer pH. Particularly, in biomedical applications, nanoparticles need to remain stable in
physiological fluids, which are highly hostile conditions due to the high ionic strength, in where a
simple surface capping is often not enough to provide colloidal stability. In order to ameliorate
this issue, engineering the interface between the nanoparticle surface and the environment is a
critical step to create functionalized and stable nanoparticles for application of interest. Hence,
stronger and covalent-like surface modulations strategies have been developed for many
biosensing applications [51, 59]. Beyond stabilizing individual nanoparticles in solution, such
surface chemistries can provide means to form self-assembled 2D and 3D nanoarchitectures in
vitro or in vivo which holds great potentials for the development of highly sensitive and selective
19

SERS nanoprobes. As explained in the previous section, forming dimeric or multimeric metal
nanostructures can indeed trigger surface plasmon coupling and create a red-shifted plasmon
resonance frequency that highly depends on the interparticle distance [45]. Therefore, precisely
controlled ligand orientation and density on nanoparticle surface becomes essential to form
uniformly separated nanoparticles, since obtaining an LSPR frequency in NIR range provides an
optimum light penetration and minimum tissue absorption for any biomedical applications.
In the last decade, Raman spectroscopy and SERS nanoparticles have gained increased
attention because of their potential use as an imaging modality for diagnostic and therapeutic
applications [60, 61]. These nanoparticles can be tailored with multiple chemical functional
groups capable of recognizing targeted moieties on cells, providing Raman signals, and
increasing the stability of nanoparticles. Moreover, Raman spectral bands are narrower than
fluorescence bands, which restrains overlapping spectral signals. Hence, this system facilitates the
development of multiplex (targeting multiple domains) detection with sharp and distinguishable
vibration bands [60, 62, 63]. The molecules that create unique Raman fingerprints are called
Raman reporters, and more than one Raman reporter functionalize the nanoparticle surface along
with different targeting ligands in the case of multiplexing [60]. Even though adsorbing the
Raman reporter on plasmonic nanoparticles enhances the inherently weak intensity of Raman
signals, localizing a Raman reporter within a hot spot further enhances the SERS signals which is
beneficial for in vivo applications.
Most inorganic nanomaterials used for biomedical applications require functional ligands
to achieve biocompatibility and solubility in physiological buffers [64]. Poly(ethylene glycol)
(PEG), which is a common surface functional group, provides water solubility, biocompatibility
and passivation against nonspecific adsorption [65, 66]. There are various lengths of the PEG
polymers that can differentially impact the colloidal characteristics of nanoparticles, so the length
20

of the PEG can be selected based on the aim of application and the presence of other chemical
groups on the same surface [67]. The capability of synthesizing PEG polymers with different
terminal functional groups makes them versatile components for nanoparticle surface chemistry
[68].
In this thesis, we have used strong thiol-metal interactions to modify nanomaterial
surfaces with PEG chains and complementary split fluorescence protein fragments. Initially, in
order to understand the surface modifications of metal nanomaterials, covalent surface
modifications of metal nanoparticles will be explained. This will be followed by the use of
Raman reporters for surface functionalization, polyethylene glycol coating and its benefits, and
hot spot formation with bio-inspired scaffolds.

1.2.2 Surface modifications
In this thesis, we have worked with plasmonic gold and silver nanoparticles (AuNP and
AgNP), therefore these metal nanomaterials will be the focus of this section. One of the most
common ways to synthesize AuNP is Turkevich method where reduction of chloroauric acid in
aqueous phase takes place [31]. In this method, citrate ions are used as reducing agents and the
produced nanoparticles possess an adsorbed layer of negatively charged citrate ions which creates
electrostatic repulsion between nanoparticles and stabilizes them in solution [69]. Citrate ions on
the AuNP surfaces form hydrogen bonds and van der Waals interactions which cannot support the
stability of nanoparticles as the ionic strength of the solution increases or in different pH
conditions [70, 71]. In high ionic strength conditions, which is the case for physiological fluids,
the electric field around nanoparticles is shielded and this can result in aggregation of
nanoparticles due to the attractive forces (i.e. hydrogen bonds, van der Waals interactions) [72].
Also, different pH conditions can induce the loss of the surface charge, or convert the sign of the
21

surface charge depending on the isoelectric point (PI), and both conditions can result in
irreversible aggregation of the nanoparticles [71]. Instead of electrostatic stabilization, strongly
binding ligand molecules can provide an improved steric stability by forming a denser protection
layer against external harsh conditions.

Figure 1.9 Schematic of different functional groups on gold nanoparticle surface and different functionalization
methods [73].

Different terminal functional groups on ligands hold different level of affinity to bind inorganic
metal surfaces through electrostatic adsorption, chemisorption, covalent binding, or non-covalent
affinity based interactions. These functional groups can be conjugated to various biomolecules,
such as DNA, enzymes, peptides, proteins, lipids, or sugars, for tailoring the nanoparticle surface
and creating hybrid materials. Metal nanoparticle/biomolecule conjugate utilizes the properties
and functions of both sides simultaneously to be used in various biomedical applications.
Thiol-metal interaction (chemisorption) is one of the most well-known covalent-like
bonds that has been used efficiently on gold and silver nanoparticles [74, 75]. In some cases, the
22

exposed cysteine residues of proteins can strongly bind to gold nanoparticle surface through thiol
groups [76, 77]. If there is a lack of thiolated residue in a native protein, functional thiol groups in
cysteine residues can also be genetically encoded into the protein sequence or conjugated to a
peptide to control the orientation on the surface [78, 79]. Thiol-conjugated single stranded DNA
linkers have been used to coat gold nanoparticles, and in the presence of complementary DNA
strand, temperature-controlled reversible aggregation of gold nanoparticles via annealing and
melting of oligonucleotides has been shown [80]. Even though, mercaptoalkyl bearing
oligonucleotides form a robust bond with gold nanoparticles, the presence of another thiol group
can be competitive and oligonucleotides can lose their functionality [81]. To overcome this issue,
steroid cyclic disulfide as an anchoring linker on gold nanoparticles have been conjugated to
oligonucleotides and resulting structures have shown higher stability and resistance to external
thiol attacks compared to the conventional single thiol bearing conjugates [81]. The anchor
groups (i.e. thiols, cyclic disulfides, or multiples cysteine residues) can easily replace the
stabilizing agents (i.e. citrate, phosphine) that are weakly adsorbed on nanoparticle surface during
the synthesis [82]. Proteins can be genetically encoded with cysteine residues which provide a
proper orientation of proteins on a gold surface. An enhanced binding ability of antibody to
antigen has been achieved due to the well-oriented protein layer on gold surfaces through the
cysteine-mediated protein immobilization [83]. In this study, N-terminus of protein G was
genetically encoded with different numbers of cysteine residues (Cys1-protein G, Cys2-protein G,
Cys3-protein G). Cysteine-tagged protein G structures have formed well-oriented layer on gold
surfaces and provided high number of human antibody (IgG) immobilization. In final step,
immobilized antibodies exhibit higher detection ability to its antigen. The orientation of the
surface ligands also depends on the coating density. When the function of a biomolecular layer is
to react with an external molecule, such as antibody-antigen interaction, it requires a sufficient
23

space to perform the reaction. As high protein coverage on metal nanoparticle surfaces might
cause steric hindrance to react with external analytes, low protein coverage can lead instability of
the nanoparticles where spacer linkers (i.e. poly(ethylene) glycol) might be required.

Figure 1.10 Different conformations of oligonucleotides on gold nanoparticles based on the surface density [84].

1.2.3 Poly(ethylene) Glycol (PEG) coating
Metal nanoparticles often suffer from becoming unstable in physiological fluids or during
the purifications steps from the excess surfactants. In biomedical applications, in vivo, possession
of long blood circulation time is a critical characteristic of nanoparticles to reach the targeted
tissues. One of the reasons of short blood circulation is the formation of protein corona
(nonspecific protein adsorption on functionalized nanoparticles) which causes a rapid uptake by
the reticuloendothelial system (RES) [85]. The protein corona around the nanoparticles is
composed of opsonins that leave antigen markers and this leads to be recognized by the immune
cells, followed internalization and clearance from the bloodstream [86]. Poly(ethylene) glycol
(PEG) polymers has dynamic conformations with different lengths of repeating ethylene ether
groups. Using PEG with an anchoring group (i.e. SH-PEG) reduces the nonspecific protein
adsorption, so called “stealth behavior”, resulting in less RES uptake and long blood circulation
time [87]. Due to the hydrophilic repeats of ethylene glycol, PEG functionalization increases the
solubility and stability in different buffers and physiologic fluids [88]. PEG functionalization also
24

improves the biocompatibility of the nanoparticles [67]. However, thiol-mediated nanoparticle
surface modulations exhibit low oxidative stability, thus sulfonates and sulfonates species are
formed upon oxidization [89]. Once sulfonate forms, it can desorb from the surface upon aqueous
rinsing or introducing fresh thiol groups [90]. Alkanethiolates are stable for a few days before
oxidization which is not sufficient to use in long term physiological applications. There are
methods to improve the long-term stability of these nanoparticles, the details are described in
ANNEX 2 (A2.1.2).

Figure 1.11 Illustrations of possible SH-PEG conformations on gold nanoparticles. (a) Number of PEG on 15 nm
AuNPs exponentially decreases as a function of the molecular weight (proportional with the length of PEG) of PEG,
inset shows the surface grafting density as a function of the molecular weight of PEG. (b) The relationship between the
number of PEG on AuNPs and the size of AuNP [67].

PEG polymers can also acquire either “mushroom” or “brush” conformation depending
on surface grafting density, as oligonucleotides do [67]. The mushroom conformation occurs
under the low surface density and long PEG chain length conditions. Whereas, the brush
conformation takes place when the surface density is high and the PEG chain length is short
(Figure 1.11a). As depicted in Figure 1.11b, the surface grafting density exponentially increase as
the nanoparticle diameter increase due to the high surface are to volume ratio of nanoparticles.
25

SH-PEG polymers can create denser and compact surface grafting to ameliorate thiol oxidation,
but at the same time they can be used as spacers between chemisorbed biomolecules (i.e. protein,
antigen, peptide) to provide enough room for functional interactions of biomolecules.
Furthermore, PEG chains supply decreased cytotoxicity and increased blood circulation through
“stealth” behavior. The facile conjugation of SH-PEG to different molecules (i.e. biotin, folate,
dopamine) makes them versatile ligands that can both anchor on metal nanoparticle surface and
carry a functional purpose (i.e. detection of receptors, sensing the presence of certain molecules)
for various applications.

1.2.4 Raman reporters
The preparation of SERS probes usually includes three components on metal nanoparticle
surface; PEG polymers, targeting biomolecules, and Raman reporters. These Raman-active
molecules have intrinsic and distinctive vibrational fingerprints. The selection of Raman reporter
and the deposition methods are critical for developing an effective SERS probe. In order to form
a stable and strong bond between the nanoparticle surface and the Raman reporter, either sulfur-
containing dyes and small thiol molecules or proteins with genetically encoded cysteine residues
and naturally cysteine-rich proteins can be selected as a reporter. Molecules with a relatively
large Raman scattering cross section provide a stronger Raman signal. Chromophores are
generally used as Raman reporters because their Raman cross sections are known to be larger
than the small molecules [91]. Using a Raman reporter with an absorption frequency that matches
the excitation wavelength at the plasmon frequency of the nanoparticles are optimal for SERRS
and can give rise to enhancement factor of 10
8
or above. This enhancement effect is significantly
higher than the SERS effect where the enhancement factor is approximately 10
6
[92]. Also, it is
important to use a Raman reporter with unique and sharp fingerprints that can be easily
26

differentiated from the environment. One of the most common sulfur-containing dyes is
malachite green isothiocyanate, which has high affinity to bind the metal nanoparticle surface,
has been used for in vivo SERS applications to detect cancer tissues [93]. Another organic dye
consists of isothiocyanate (-N=C=S) group is TRITC (tetramethylrhodamine-5-isothiocyanate)
also shows high affinity to metal surface, however its Raman vibrational bands are active in
Uv/Vis range, hence it has less potential use in in vivo applications [94]. DTTC (3,3ʹ-
diethylthiadicarbocyanine iodide) is also a sulfur-containing, but a cyanine derivative dye with a
large Raman scattering cross section and NIR absorption frequency that makes it versatile Raman
reporter for in vivo applications (Figure 1.12) [93]. On the other hand, small organic molecules
with thiol, such as 4-aminothiophenol, 2-naphthalenethiol, have small Raman scattering cross
section, although they have strong binding on metal surfaces and they are suitable for
multiplexing [95]. Since these molecules are small, large number of surface grafting can be used
to obtain strong Raman signal [96]. However, it has been reported that above a certain
concentration of 2-naphthalenethiol has cytotoxic effect on cells [97]. There are other Raman
reporter molecules with strong and sensitive Raman signals, such as crystal violet, rhodamine B
and rhodamine 6G, although these reporters interact with the metal surface through
electrostatically forces or amine-metal binding where the interaction is not as strong compared to
thiol-metal bonds. Therefore, these dyes suffer from weak affinity and weak signal stability [95].
27

Figure 1.12 SERS spectra of two different Raman reporters. (a) DTTC functionalized gold nanoparticles are applied in
vivo mouse model, and SERS signals are detected from different locations in animal; subcutaneous injection spectrum
(green), deep tissue injection spectrum (blue), as controls pure tag spectrum (red) and skin spectrum (black) [93]. (b)
SERS spectra of 2-naphthalethiol on silver nanoparticle embedded beads [96].

In the last decade, fluorescent proteins have been gaining attention for SERS studies. The
green fluorescent protein (GFP) from Aequorea Victoria is one of the most common fluorescent
proteins in biological application. The most critical characteristic of this protein is the
autocatalytically formed chromophore inside the beta-barrel shaped protein (4nm x 2nm) [98].
Three aminoacid residues, Ser65-Tyr66-Gly67, form the chromophore which is stabilized by
complex hydrogen bonding system [99]. Once the chromophore is formed, GFP protein gains
unique Raman fingerprints that can be utilized in SERS applications at single molecule sensitivity
[100]. Absorption frequency of GFP is also in UV/Vis range, however spectral variants of the
GFP structure can be produced by introducing site specific mutations on the protein sequence
[101]. Comprehensive characterization of GFP and complementary split fluorescence protein
fragments are detailed in CHAPTER 2. In order to attach to metal surface, either cysteine
residues can be encoded to N- or C-terminus of the protein or existing cysteine residues in the
28

structures can be used. These features make GFP a versatile Raman reporter for SERS
applications both in vivo and in vitro.

Figure 1.13 Surface coating strategies for SERS probes with (a) denatured BSA, (b) SH-PEG, (c) amphiphilic diblock
copolymer, (d) liposome and (e) silica shells [95].

Once these Raman reporters attach to the surface, they require an encapsulation layer to
prevent the ligand dissociation (Figure 1.13). As mentioned in the previous section, PEG and
block copolymers are commonly used to form a protection layer. Liposome encapsulation with
different Raman reporters has also been shown as an effective strategy to control the surface
modification (Figure 1.14) [102]. Silica coating is another encapsulation method that protects the
functionalized surface from desorption or nonspecific adsorption, provides water solubility and
stability. However, the intensity of the Raman signal dramatically decreases due to the silica
layer, therefore this strategy yields less sensitive SERS probes [103]. This type of Raman probes
is called nanotags and has been used widely in biomedical applications because of their stable
structures.
29

Figure 1.14 Liposome encapsulated SERS nanoprobe. (a) Chemical structure of malachite green isothiocyanate
(MGITC). (b) TEM image of a bilayer-encapsulated and MGITC functionalized gold nanoparticle, light gray ring
around the nanoparticle represents bilayer coating. (c) SERS spectrum of MGITC on gold nanoparticles encapsulated
with liposome [102].

1.2.5 Hot spot formation of metal nanoparticles
The surface design of nanoparticles is a critical step to maximize the electromagnetic
enhancement and optimize the efficiency of the system. When metal nanoparticles are in close
proximity, the LSPR frequency shifts due to the interparticle plasmon coupling effect which
highly depends on the gap distance between nanoparticles [45, 104]. This interparticle space is
called “hot spot” and has been well-known as “favorable position” for massive SERS
enhancement of Raman reporters [105, 106]. Formation of hot spots has a major contribution on
electromagnetic enhancement to observe single molecule SERS (SMSERS) [107]. Different
bottom-up nanofabrication methods of colloidal metal nanoparticles have been developed to
create hot spots. In the first SMSERS experiments, simple alterations on electrostatic interactions
have been used to create aggregated structures [4, 108]. In these studies, after surface
functionalization with a Raman reporter (rhodamine 6G or crystal violet have been used),
increased salt concentration has driven the formation of aggregates. However, aggregation
created by salt concentration causes randomly formed, polydispersed, irreproducible
nanoaggregates which are hard to analyze. To ameliorate this issue, silicon oxide (SiO 2)
encapsulation has been used on Raman reporter adsorbed metal nanoparticle assemblies and field
30

flow fractionation method has been applied to decrease the structural diversity within the sample
(Figure 1.15) [109]. Even though the hot spot formation method is nonspecific, silica
encapsulation provides a better consistency in SERS signal.

Figure 1.15 SiO2 encapsulation on PCEPE functionalized gold nanostructures. (a) Transmission electron microcopy
image of three gold nanoparticles functionalized with the Raman reporter, PCEPE, and encapsulated with silica shell
(gray ring). (b) LSPR frequencies of L-shaped nanostructure measured by dark-field Rayleigh scattering microscopy.
(c) SERS spectrum of L-shaped nanostructure. (d) Raman spectrum of PCEPE molecule. (e) Calculated Raman
spectrum of PCEPE and its chemical structure [109].

Modulation of salt concentration gives rise to a nonspecific hot spot formation, on the
other hand, the modification of gold nanoparticles with thiol functionalized oligonucleotides has
brought up a unique fashion on controllable hot spot formation (Figure 1.16) [80].
31

Figure 1.16 DNA hybridization on gold nanoparticles. Thiol-conjugated oligonucleotides functionalize gold
nanoparticle surface. Upon introducing linking DNA duplex, temperature dependent, reversible oligomerization creates
controllable gold nanoparticle aggregates [80].

In this method, two gold nanoparticle groups have been functionalized with thiol-
conjugated oligonucleotides that can hybridize upon linking DNA duplex introduction to the
system. These highly controlled 3D nanoaggregates are extremely stable against high salt
concentration and elevated temperatures. At high temperatures, DNA duplex dehybridizes and
gold nanoparticles disassemble. Whereas at low temperatures, DNA strands starts to anneal and
gold nanoparticles creates assembled structures. Since the length of the DNA describes the
distance between nanoparticles, this system provides highly controllable nanogap distances. Also,
hot spot formation becomes highly specific and controllably irreversible compared to salt driven
aggregation, due to extreme specificity of DNA hybridization reaction. The self-assembling DNA
scaffolds has been widely used to form hot spots and detect multiple biomolecules in SERS
32

studies. In order to create uniform 1 nm gap distance within the sample, a new gold nanostructure
called gold nanobridged nanogap particles (Au-NNPs) have been developed (Figure 1.17) [110].
In this system, DNA-tailored AuNPs have been used as seeds to synthesize Au-NNPs along with
polymer, reductant, and precursor addition. The second gold layer has been completely formed
depending on the concentration of reagents. DNA provides a versatile platform to precisely
localize the Raman dye, ROX (carboxy-X-rodamine). Therefore, the Raman dye has been
localized into the uniform and well-defined 1 nm gap with large surface area. These robust
nanostructures have high potential to use for biosensing and imaging applications.

Figure 1.17 Synthesis and characterization of gold nanobridged nanogap particles (Au-NNPs). (a) Scheme of forming
gold layer on the template DNA-modified gold nanoparticles. (b) TEM images of intermediates and fully synthesized
Au-NNP. (c) Comparison between uncontrollable nanoaggregates with non-uniform nanogap size that cause non-
uniform SERS signal (left) and controllable nanostructures with well-defined nanogap that yield robust and quantitative
SERS signal (right) [110].

Alternatively, complementary protein and peptide fragments offer many advantages for
the controlled self-assembly of metal nanoparticles. These complementary fragments have high
33

chemical specificity and binding affinity. Amongst complementary protein structures, split
fluorescent proteins are the ones that can be genetically programmed to rapidly self-assemble and
serve as a molecular glue between nanoparticles. Their folding mechanism into a beta-barrel
shaped structures is well defined and these biocompatible split protein fragments form a
chromophore inside the barrel structures upon complementation [111]. In this system, since the
autocatalytically formed chromophores carry unique Raman fingerprints, split fluorescent protein
fragments become activatable Raman reporters upon complementation [112, 113]. When metal
nanoparticles are functionalized with split fluorescent protein fragment and the complementary
peptide fragment, the complementation of these two fragments drive the self-assembly of
nanoparticles into photonically active SERS hot spots. The self-assembly process can occur in
cells and this method has a great potential for in vivo biomedical applications because i) these
hybrid nanoprobes are biocompatible, ii) functionalized fragments have high affinity to
complement each other, iii) the Raman reporter become activated in situ, iv) compared to the
oligonucleotide building blocks, these fragments are not sensitive to nucleases and do not depend
on a linking DNA duplex component.

1.2.6 Conclusion
In this chapter, surface functionalization methods for metal nanoparticles were
summarized. Molecule-metal surface interactions were mostly focused on covalent-like thiol-
metal surface bonding and its comparison with nonvalent bonds; electrostatic adsorption,
hydrogen bonding, and van der Waals forces. Even though thiol groups form a strong binding
with metal surfaces, the surface grafting density plays an important role in stability of the
nanoparticles against harsh conditions. To overcome this issue, poly(ethylene) glycol polymers
come into play and not only form a brush layer on the surface to increase the stability, but also
34

contribute in many different angles, such as increased water solubility and biocompatibility,
surface passivation against nonspecific opsonin adsorption by stealth behavior, hence decreased
immune system attack and clearance of the nanoparticles, creating a platform to conjugate
recognition molecules for certain receptors. Furthermore, the importance of Raman reporter with
peculiar Raman bands was emphasized in order to build a SERS probe. Sulfur-containing dyes,
small organic molecules with thiol groups, proteins with either native cysteine residues or
genetically encoded cysteine residues are well-defined Raman reporter candidates. Also, green
fluorescent protein and its spectral variants hold a great potential for single molecule SERS
studies. These fluorescent proteins will be analyzed thoroughly in CHAPTER 2. Each of the
surface components of the SERS probes carries a significance to improve the electromagnetic
enhancement. Different methods have been explained to create hot spots which provides further
increase in the electromagnetic enhancement. For in vivo biomedical applications, in situ
formation of hot spots is clearly a challenging task in where split fluorescent fragments offer a
promising approach to ameliorate the issues comes with physiological environments. Hot spot
formation with split fluorescent fragments will be explained in depth in CHAPTER 3.
The surface chemistry of metal nanoparticles was explained under three main subtitles;
(i) the ligand-surface interaction methods, (ii) the functional groups to develop stable nanoprobes,
(iii) the strategies for hot spot formation. The titles summarized in this section are essential to
understand the purpose of developing different variants of split-FPs (CHAPTER 2). Moreover,
the methods explained in this section will help to understand the development of SERS probes
where autocatalytic cyclization of split-FP fragments drive nanocluster formation with
photonically active hot spots (CHAPTER 3). In the next section, the application platforms of
SERS nanoprobes will be discussed.

35

Biomedical Applications of Sers Probes
1.3.1 Introduction
Metal nanoparticles possess versatile properties, such as high surface area to volume
ratio, facile synthesis methods of different structural features, available surface modification sites,
ability to confine and amplify the electromagnetic field through collective oscillations of surface
plasmons, tunable LSPR frequency and programmable self-assembly process as explained in the
previous chapters. These properties make metal nanoparticle suitable components for wide
variety of applications, for instance SERS [110], non-linear optics [114], superlattice
mesostructures [40], metamaterials [115], photothermal transduction [116] or hot electron-hole
pair generation [117]. Certainly, biosensing and biodetection methods with metal nanoparticles
have also attracted enormous attention due to purpose-oriented development. The LSPR shift of
DNA-tailored gold nanoparticles upon hybridization with the targeted DNA strand triggered the
development of colorimetric biomolecule detection methods [80]. Furthermore, these
nanoparticles have been used for oligonucleotide arrays to detect particular DNA sequences at a
very high sensitivity levels (<100 pM) which has been further increased up to 50 fM by silver ion
reduction on gold nanoparticles [118]. Modifying gold nanoparticle surface with Raman reporter
conjugated oligonucleotides has allowed multiplexed analyte detection owing to the spectral
signatures at 1 fM sensitivity [119]. Developing stable, water soluble, and biocompatible SERS
nanoprobes has encouraged the in vivo applications, particularly the detection of cancer tissue.
Raman reporter coated gold nanoparticles have been encapsulated with silica shells that are
attached to PEG and EGFR binding affibodies, and after the intravenous injection of the targeted
SERS probes, the multiplexed SERS signal has been detected from both deep and superficial
tissues [120]. Furthermore, SERS nanoprobes have been used for photothermal therapy in cancer
cells in vitro and in vivo where upon laser illumination, metal nanoparticles create heat radiation
36

that can kill the targeted cancer cells [121]. Due to these multifunctional properties of SERS
nanoprobes, they create an exceptional platform for theranostics.
In this section, the three major biomedical application groups will be described as
biosensing, biomedical imaging, and therapeutic/theranostic applications. In this thesis, we used
SERS imaging and photoacoustic microscopy, therefore the focus of this section will be on SERS
based biosensing, SERS imaging, photoacoustic imaging, and photothermal therapy. More details
about biomedical applications of SERS nanoprobes can be found in ANNEX 3.

1.3.2 SERS nanoprobes for biosensing
Multifunctional plasmonic nanoprobes deliver ultrasensitive detection of various disease
biomarkers owing to their highly tunable size, shape, LSPR frequency, and large surface area for
versatile surface modifications. LSPR frequencies of nanoparticles due to their different sizes and
geometries provide LSPR-based label-free in vitro detection of multiplexed analytes. There are
various applications that take advantage of metal nanoparticle properties to detect biomolecules,
such as MicroRNAs (miRs), single stranded and non-coding RNAs, that are key regulators for
various biological processes [122]. Altered miR expression is an important indicator for early
diagnosis of pancreatic ductal adenocarcinoma [123]. Functionalized gold nanoprisms with PEG
and complementary single stranded DNA (ssDNA) has shown an LSPR shift upon hybridization
with miRs and the system has provided a high detection sensitivity compared to the conventional
methods (Figure 1.18a) [124]. As mentioned in the previous section, highly tunable hot spot size
is a critical characteristic for SERS-based nanoprobes. La Rocca et al. has reported the use of 3D
vertical plasmonic nanoantennas to chemically analyze the microenvironment of living cells
(Figure 1.18b) [125]. In this system, silver/gold bilayer nanoantennas were produced with inner
silver core providing high SERS enhancement factor and thin gold layer increasing the
37

biocompatibility and stability. Neuroblastoma (N2a) and fibroblast (NIH/3T3) cells grown on
nanoantenna arrays were used to acquire detailed SERS spectra with NIR excitation source at low
powers without damaging the live cells. Furthermore, reduced graphene oxide (rGO) has been
used as a separator between plasmonic gold (Au) and silver (Ag) materials, hence narrow gap hot
spots has been formed (Figure 1.18c). The distinct Raman signals of graphene, D and G bands,
has been distorted upon tumor cell growth on these substrates and the distortion has been shown
as tumor cell specific [126].

Figure 1.18 Plasmonic biosensor design based on LSPR frequency, shape, and material. (a) Gold nanoprisms modified
with PEG and single stranded DNA (ssDNA) exhibit extinction spectrum shift upon miR-X incubation in PBS buffer
[124]. (b) SEM image of vertical plasmonic nanoantenna (left), representative drawings of cell/nanoantenna (center)
and cell/planar SERS substrate (right) [125]. (c) Schematic of reduced graphene oxide (rGO) separator between Au and
Ag nanoclusters on silicon (Si) substrate. Change in Raman spectra when normal (202) and tumor (7402) cells were
grown on rGO substrates; Raman spectra of rGO on Si slide (green dash line, Ag@rGO@Au), normal live cells on
38

rGO substrate (blue solid line, 202@Ag@rGO@Au), and tumor cells on rGO substrate (red dash line,
7402@Ag@rGO@Au) [126].

SERS based biosensing techniques have a wide range of applications in real-time
monitoring of chemical interactions in microenvironments, diagnosis of early cancer stages, and
detection of diseases through antibody-antigen interaction at a significantly high sensitivity.
Therefore, these systems hold a great potential for clinical diagnostics.

1.3.3 SERS nanoprobes for biomedical imaging
Cellular level monitoring and mapping of different biological processes, such as cell
proliferation, intercellular interactions, biochemical signal transduction, impaired protein
expression and apoptosis, etc., has gained a great importance due to the lack of sensitivity and
resolution of conventional methods. X-ray imagin, computed tomography (CT) or positron
emission tomography (PET) use radioactive waves and toxic chemicals as contrast agents. MRI
(magnetic resonance imaging) is also another conventional method for diagnostic purposes and
determining the stage of the disease with radio waves, strong magnetic fields and field gradients.
Even though MRI can be seen as superior to CT because MRI does not involve x-rays and it is
less hazardous compared to CT, each of these methods provides different diagnostic information.
None of these anatomical and metabolic imaging systems alone are sufficient to detect early stage
diseases or to remove a tumor tissue precisely, since they suffer from the limited spatial
resolution of a few millimeters to hundreds of micrometers. Coupling these imaging tools is a
common strategy to overcome the deficiencies of each individual technique. To overcome the
lack of spatial resolution, Kircher et al. has combined MRI, photoacoustic imaging and SERS
imaging [127]. They have used MRI to define the macroscopic margins of tumor tissue, however
this technique has limited spatial resolution, and cannot provide the exact tumor borders.
39

Photoacoustic (PA) imaging can improve the depth and the resolution limits. In PA imaging,
targeted molecules create heat and thermal expansion upon interacting with light pulses and the
resulting ultrasound waves can be recorded by a transducer [128]. Both MRI and PA imaging
require a contrast agent that should retain in tumor tissue long enough to define the tumor
margins pre-surgery, and also during the surgery to remove the tumor tissue precisely. SERS
nanoprobes are good candidates to be multimodal contrast agents for both MRI, PA imaging, and
also for SERS imaging. De La Zerda et al., have produced multi-layer gold nanoparticle probes.
First, gold nanoparticles were functionalized with a Raman reporter (trans-1,2-bis(4-pyridyl)-
ethylene) for SERS imaging; second, silica shell was used to encapsulate the Raman reporters;
third, a thin gold layer was coated on silica shells; fourth, the MRI contrast reagent DOTA-Gd
3+

was appended on the thin gold layer. Due to the high optical absorbance characteristics of gold
nanoparticles, these probes increase the heat production and thermal expansion significantly,
hence they improve the PA imaging sensitivity [129]. After the injection of the SERS probes, the
clear visualization of the tumor tissue in the live mouse brain was obtained through three different
modalities that are colocalize with each other. Tumor resection has also been carried out using the
PA and SERS imaging guidance which provide the detection of small cancerous foci locations.
Photoacoustic imaging has been shown as a noninvasive and versatile tool for wide range
of biological tissue applications, such as detection of in vivo brain lesions [130], human breast
tumor-related blood vessel deformations [131], human skin microvasculature networks [132]. PA
imaging employs the conversion of optical excitation into ultrasonic detection [133]. The
biological tissues absorb the nanosecond-pulsed laser beam, consequently confined energy gives
rise to heat production (ΔT). Subsequently, the tissues undergo thermoelastic expansion that
generates acoustic waves and creates pressure difference (p). It can be written as;
𝑝 = 𝛽 ∆𝑇 /𝜅 1.6
40

where β is the thermal expansion coefficient and к is the isothermal compressibility. Since β is
approximately 4x10
-4
K
-1
and к is 5x10
-10
Pa
-1
, a 1 mK temperature difference creates a 800 Pa
pressure rise which is ultrasonically detectable because this pressure difference is well above the
noise level of an ultrasonic transducer [134]. Therefore, photoacoustic imaging delivers high
signal-to-noise ratio without thermal damage on the tissue [135]. The change in photoacoustic
pressure depends on the thermal and mechanical characteristics of the tissue and the optical
energy deposition [133]. In tissue, the ultrasonic scattering coefficient is approximately 2-3 orders
of magnitudes smaller than the optical scattering coefficient. Hence, detection of ultrasonic waves
provides higher spatial resolution in tissue than the optical imaging [136]. Furthermore, the
center frequency and the bandwidth of the ultrasonic detection system are important parameters
to define the spatial resolution. In the case of high center frequency (i.e. 300 MHz) and broad
bandwidth, the spatial resolution increases whereas the penetration depth decreases [133, 137]. In
PA imaging, visible and NIR wavelengths can be used for optical irradiation of the tissue where
NIR range provides several centimeters penetration depth [135]. Different molecules in a tissue
exhibit different optical absorption behaviors which can provide functional imaging to describe a
physiological map. From this standpoint, PA imaging is highly sensitive for oxy and
deoxyhemoglobin, and melanin molecules at NIR range. On the contrary, water and lipids are not
primary contrast sources in this range [132]. Cancer tissues have peculiar structures in terms of
the density of the capillaries which is significantly higher than the normal tissues, with distorted,
irregular vessel systems. Consequently, these two peculiarities give rise to an increased
hemoglobin concentration in tumor tissues, and a strong photoacoustic signal [138]. However,
tumor structures consist of many single cells and the number of these cancerous cells is a critical
parameter to detect the early stage of the tumor [139]. This is a limiting factor for many imaging
tools, for instance, since the size of a single cancer cell is approximately 10 μm, these cells do not
41

show strong intrinsic absorption contrast. To overcome this issue, extrinsic optical contrast agents
can be implemented to enhance the absorption and hence photoacoustic signal [133].
Plasmonic metal nanoparticles become a promising contrast agents for PA imaging due to
their high optical absorption and light scattering cross section compared to other absorbing
molecules and organic dyes [140]. Particularly, biocompatible gold nanostructures have been
widely used to detect and diagnose various cancer tissues [141, 142]. Earlier, non-targeted gold
nanocages have been functionalized with PEG and applied to rat cerebral cortex [141]. In this
study, the enhanced permeability and retention (EPR) effect of tumor regions allowed the
accumulation of non-targeted gold nanocages at the tumor tissue. Even though this leaky
vasculature in tumor regions is an advantage for non-targeted nanoparticles, relatively big size of
gold nanoparticles can be detected by the reticuloendothelial system (RES). For this reason,
single-walled carbon nanotubes (SWNT) have been used as contrast agents for PA imaging
(Figure 1.19) [128]. In this study, 1-2 nm in diameter and 50-300 nm in length SWNTs have been
grafted with RGD peptide conjugated phospholipid-PEG 5000 (PL-PEG 5000). RGD peptide
recognizes integrins, particularly α vβ 3, over-expressed in tumor vasculature.
42

Figure 1.19 Single-walled carbon nanotube (SWNT) application on tumor bearing live mice. White dotted line shows
the vertical slice through the tumor. Ultrasound (grey) and photoacoustic (green) images were overlapped at pre- and
post-injection time points. In order to observe the effect of the SWNTs, pre-injection images were subtracted from
post-injection images to create subtraction images. The high photoacoustic signal (white arrows) in plain SWNT
applied sample was not observed in the subtraction image which indicates the presence of large blood vessels.
However, RGD peptide grafted SWNT applied sample shows high photoacoustic signal in the subtraction image which
shows the presence of targeted SWNTs in the tumor region [128].

However, the toxicity of SWNTs is highly debatable. Some studies have shown no
toxicity of SWNTs in vivo [143, 144], yet some have reported significant toxic effect of these
contrast agents [145, 146]. Compared to carbon nanotubes, well-functionalized gold
nanostructures do not exert high cytotoxicity either on in vitro cell culture or in vivo animal
models [147]. Therefore, highly tunable size, large modifiable surface, and lack of toxicity make
gold nanoparticles ideal contrast agents for PA imaging. Quantitative analysis of PA signal of
gold nanoparticles in prostate cancer cells has shown that the controllable surface
functionalization carries a great importance to optimize the PA imaging [148]. In this study, gold
43

nanoparticles (30 nm in diameter) have been functionalized with PEG and RGD peptides. The
increased concentration of RGD peptide on the gold nanoparticle surface improves the cellular
recognition and uptake (Figure 1.20b). Even though increased cellular uptake enhances the PA
signal, this effect halts and reverses the PA signal intensity at above certain concentration level
(Figure 1.20a). This phenomenon reflects the effect of LSPR red shift due to the higher levels
intracellular aggregation.

Figure 1.20 In vitro and in vivo photoacoustic imaging with gold nanoparticle as contrast reagents. (a) Normalized
photoacoustic signal intensity of 2D photoacoustic images with respect to the RGD peptide surface density.
Photoacoustic signal intensity increases up to 1000 RGD peptide density, then the intensity decreases with increased
surface density. (b) Bright field and fluorescence microscopy images of human prostate cancer cells (A-C). Three
photoacoustic microscopy (PAM) images show single cancer cells in red dashed circles (D-F). Scale bar: 100 μm,
close-up scale bar: 25 μm [148].

Raman microscopy systems for in vitro cell culture or in vivo live animal applications
employ a focused laser spot for 2D spatial scanning and either 1D or 2D array CCD. Systems
equipped with 2D array CCD deliver y-axis information with corresponding Raman spectrum.
44

When the results are calibrated with predefined x-axis dimension, 2D spatial scan with
corresponding Raman spectra over a sample area can be achieved. Focused laser point scanning
maximizes the excitation power density on defined sample area, however it has limited capability
when a wide sampling area is selected to image. Depending on the resolution of the laser spot
movement, obtaining a reconstructed image from a few hundred micrometer square area, hours
long acquisition times are required which is not optimum for in vitro cell cultures due to the cell
death, and the cell movements over time and is not even applicable to in vivo animal imaging due
to the limited anesthesia duration. Shorter integration time can be used to decrease the image
acquisition time, but this also decreases the spatial resolution. To improve the scanning time, a
cylindrical lens can be implemented into the system to create a laser line out of a laser spot, hence
a bigger area can be scanned compared to point scanning [149]. Even though this will decrease
the full scanning time, it will also decrease the laser power density over a line. This system is
well-adapted for small area in vitro live-cell Raman imaging where the laser line is fixed, but the
stage provides the scanning capability. To further improve this system, 2D galvanometric mirrors
can be attached to the light path which facilitate the rapid laser scan on a fixed stage which is
favorable for in vivo animal imaging to ease the translation of the animal (Figure 1.21a) [120].
Here, four different SERS nanotags have been subcutaneously injected in living mice torso.
Multiplexed nanoparticles have been detected ten times faster than a system without
galvanometric mirrors. Moreover, when these SERS nanotags intravenously injected to a mouse,
the system was still able to achieve multiplexed detection of the nanotags in the liver due to RES
clearance. It has been shown that this raster scanning with galvanometric mirrors provides an
order of magnitude faster acquisition time with equivalent sensitivity, spectral/spatial resolution,
and multiplexing capability. This system is called raster scanning/mapping resulting in a
hyperspectral cube of subsequent planes where x-axis corresponds to the spectral dimension
45

(Raman shifts), y-axis represents spatial y-axis coordinates that scanned along the laser line, and
z-axis represents each plane that scanned through the spatial x-axis coordinates. After data
acquisition, the entire image can be reconstructed at a single Raman shift of interest or as an
integration of a range of Raman shifts.

Figure 1.21 Layout of optical systems. (a) 2D galvanometric mirror facilitated rapid laser scanning system for Raman
imaging [120]. (b) Hyperspectral Raman imaging system with laser beam shaping module [150].

Furthermore, a SERS imaging system without a cylindrical lens has been used for
multiplexed detection of subcellular organelle-targeted gold nanoparticles in live cells (Figure
1.22) [151]. In this study, previously described gold nanobridged nanogap particles (Au-NNPs)
have been functionalized with both Raman reporters and targeting moieties. Three groups of
nanoparticles have been functionalized with both Raman reporters; MB, 44-DP, AB at the narrow
intra-nanogap region and targeting moieties; RGD peptide for cytoplasm, MLS-RGD peptide
combination for mitochondria, NLS-RGD peptide combination for nucleus targeting on the
surface of nanoparticles, respectively. Since these Raman reporters are NIR sensitive, they are
46

compatible for biological applications. Also, locating the Raman reporter in an intra-nanogap
provides a uniform Raman signal in solution. Besides these advantages, the confocal SERS
imaging system employs galvanometric mirrors to allow fast scan on fixed, stable stage and
provides high resolution single live cell Raman imaging in less than a minute.

Figure 1.22 Subcellular organelle targeting with gold nanobridged nanogap particles (Au-NNPs). MB Raman reporter
was used to modify Au-NNPs along with mPEG-thiol and RGD peptide to target the cytoplasm; 44DP Raman reporter
was used to modify Au-NNPs along with mPEG-thiol, RGD and MLS to target mitochondria; AB Raman reporter was
used to functionalize the nucleus targeting Au-NNPs with mPEG-thiol, RGD and NLS peptides. Representative Raman
spectra from inner cellular region at 3 h, 6 h, 12 h time points. Bright field and Raman images were overlaid at 0 h, 3 h,
6 h, and 12 h time points [151].

In SERS imaging of both in vitro and in vivo models, the Raman fingerprint of the
reporter is the most prominent Raman shift among the rest of the peaks in the spectrum. In order
to acquire only one Raman shift, angle-tunable bandpass filters are implemented into a system
where the defocused laser beam (ca. 100 μm) allows the Raman imaging without multiple
scanning steps [152-154]. Due to the angle-tunable filters, this setup can avoid the grating
component of the regular Raman systems. Multiple scanning steps can acquire different Raman
47

band signals from the same region of the sample. This system is called global imaging/mapping
resulting in a hyperspectral cube of subsequent planes where x- and y-axes corresponds to the
spatial dimensions and z-axis represents the spectral dimension where different Raman band
signals from the same region of interest exist, and the series of z-axis planes forms the spectral
range. To eliminate the non-homogenous laser illumination over the field of view due to the
defocused laser beam, a laser beam shaping module has been used (Figure 1.21b) [150]. The
reshaped laser beam can be sent to the microscope objective through a dichroic mirror and
collected and collimated with the same objective. The emitted light passes through the dichroic
mirror to suppress the Rayleigh scattering and to transmit Raman scattered light. This Raman
scattered light can be directed to the Bragg tunable filters where a single resonant Raman band is
diffracted, and the rest of the Raman bands are refracted from the optical path. The diffracted
Raman band can be sent to the CCD by a mirror and a plane of an image can be created. Using
this system, yeasts cells, Candida albicans, tagged with PEG modified SWNTs that encapsulates
a Raman reporter, β-carotene (βcar), have been imaged at 1520 cm
-1
Raman band [155].
Furthermore, Gaufres et al. have shown a high resolution image of biotinylated PEG-SWNT-βcar
on microcontact printing immobilized streptavidin substrates. Therefore, this setup also provides
high resolution hyperspectral Raman images of biological materials.

1.3.4 SERS nanoprobes for therapeutic and theranostic applications
Conventional cancer treatments employ surgical removal of the tumor tissue,
chemotherapy, or radiotherapy [156]. Surgical removal techniques suffer from both the
accessibility of the tumor tissue and the sensitivity to detect and remove all the cancerous cells.
These issues can be overcome by the theranostic properties of the plasmonic nanoparticles that
involve both diagnostic and therapeutic applications. Furthermore, chemotherapy suffers from the
48

side effects on healthy tissues as well as invasive radiotherapy effects. Therapeutic properties of
plasmonic nanoparticles can be also used to ameliorate the dramatic side effects of the traditional
methods.
Facile surface modification of plasmonic nanoparticles makes them suitable carriers for
drug delivery systems. Cargo molecules can be loaded onto nanocarriers through covalent bonds,
covalent-like bonds (gold-sulfur), and non-covalent bonds (hydrogen bonding, electrostatic
interactions, van der Waals forces). Tumor tissue can be detected by actively or passively targeted
loaded nanocarriers which can be controlled remotely to release the drug [157]. Also, the drug
release process can be triggered by the biochemical properties of the microenvironment. Gold
nanorods (AuNRs) have been used for directional modifications of targeting moieties and drug
carrier gold nanoparticles (AuNP) [158]. In this study, CEM, a nondrug resistant leukemia cell
line that overexpresses protein tyrosine kinase 7 (PTK7) on the cell membrane has been used as a
model system with a control Ramos lymphoma cell line. AuNRs have been modified on the end
faces with an aptamer, Sgc8, that specifically binds to PTK7. Doxorubicin (Dox) cancer drug has
been conjugated to drug loading/anchoring DNA strands and the conjugate functionalizes the
small AuNP (8.5 nm in diameter) surface. Dox loaded AuNPs self-assemble on the side faces of
AuNRs through DNA hybridization. PEG has also been used for biocompatibility purposes.
Selective detection of CEM cells via targeted- and drug-loaded AuNP/AuNR nanocomplexes
have been followed by the internalization of the nanocomplexes. Drug-loaded AuNPs have been
released from the AuNR surface by NIR irradiation. Although significant cell death has been
reported without NIR irradiation due to the cytotoxicity of Dox, further cell death has been
achieved with NIR irradiation. This effect has been attributed to NIR-activated nuclear
accumulation of small AuNPs (8.5 nm) instead of photothermal effect of NIR irradiation, since
there has been no cell viability difference between 15 nm and 8.5 nm Dox-AuNP applied cells
49

without NIR irradiation. Therefore, actively targeted and drug-loaded nanocomplexes have shown
an enhanced therapeutic efficacy with NIR irradiation drug release mechanism. On the other
hand, passive tumor targeting has been shown with tumor-bearing mice model and paclitaxel
(PTX) cancer drug functionalized gold nanoparticles [159]. Here, PTX drug has been conjugated
to PEG molecules that are appended to AuNP (3-4 nm in diameter) surfaces. PTX-PEG-AuNPs
have shown long blood circulation time which increases the passive tumor accumulation through
EPR effect. The controlled release of PTX drug upon a dual simultaneous stimulation (using
glutathione and pig liver esterase) has exhibited a better tumor necrosis compared to the direct
administration of the Taxol® drug. These results suggest nanocarriers have a potential use to
increase the blood circulation and passive tumor accumulation due to the angiogenesis and higher
concentration of drug release on tumor site.
Theranostic nanoprobes integrate the diagnostic biomedical imaging capability and the
treatment of cancerous tissues. Besides being a nanocarrier in drug delivery systems, these
multifunctional nanoprobes contribute to image-guided photothermal therapy in vitro and in vivo.
When nanoprobes are irradiated at LSPR frequency, excited electrons couple with phonons and
this leads to heat production and thermal expansion. Plasmonic nanoparticles possess high
absorption cross section than organic or inorganic photoabsorbers, such as indocynanine green or
iron oxide, respectively (Figure 1.23) [160]. Tuning the nanoparticles to create strong absorbers at
NIR range, where the light has maximum depth of penetration in the tissue due to low scattering
and absorption of intrinsic chromophores, is critical to apply a therapeutic dose that can induce
hyperthermia on tumor tissue [161]. Under hyperthermia conditions, passively or actively
targeted tumor tissues undergo cell membrane and protein damage resulting in cell death [162].
Moreover, cancer cells are more prone to apoptosis under elevated temperatures due to the
hypoxic microenvironment compared to the healthy cells [163]. Nanoshells functionalized with
50

PEG have been applied to human breast epithelial cells (SKBr3) in vitro and transmissible
venereal tumor-bearing mice in vivo [160]. In vitro cell culture treated with nanoshells and NIR
(~800 nm) irradiation has been exposed to three different staining methods to analyze cell
viability, membrane permeability, and nanoshells binding (Figure 1.23a). In order to confirm cell
viability, calcein AM has been used which emits green fluorescence if the live cells are stained.
Only NIR irradiated sample has shown green fluorescence emission over the entire sample,
however nanoshells applied and NIR irradiated sample has not emitted fluorescence from the area
overlaps with the laser spot. Also, cell permeability has been tested by fluorescein-dextran
staining which cannot react with intact cell membrane. Nanoshell/NIR treated cells have shown
the fluorescein dextran stain over the area matches the laser spot, whereas no staining has been
observed from only NIR irradiated cells. By silver staining, cellular binding of nanoshells has
also been characterized. Therefore, in vitro thermal ablation driven by nanoshells/NIR couple has
been clearly exhibited. Furthermore, in this study, it has been shown that significantly less laser
dosage has been sufficient to induce irreversible tissue damage than the experiments used
indocynanine green as a thermal coupling agent. Nanoshell/NIR applied tissue has shown positive
silver staining and negative hematoxylin staining on the thermally ablated region (Figure 1.23b).
This has proven the presence of nanoshells and absence of nuclear structure of the cells. The good
correlation between the gross pathology results of the excised tumor tissue, silver staining,
hematoxylin/eosin staining and real-time magnetic resonance temperature imaging (MRTI)
suggests the potential use of nanoshells/NIR treatment for photothermal ablation of in vivo tumor
tissues.
51

Figure 1.23 The potential use of nanoshells/NIR treatment for photothermal ablation of in vivo tumor tissues. (a)
Calcein AM stained cells show cell viability after the NIR laser treatment. However, nanoshells/NIR laser combination
creates a clear cell death. Fluorescein dextran staining is used to analyze the membrane integrity. Nanoshell/NIR laser
combination damages the cell membrane integrity resulting in lack of cellular uptake of fluorescein dextran. However,
only NIR laser treatment does not cause enough damage on the membrane, thus due to the cellular uptake fluorescence
signal cannot be observed. (b) Histological analysis on in vivo nanoshells/NIR laser treated tumor (transmissible
venereal tumor (TVT)) bearing mouse model. (1) Gross pathology after the nanoshells/NIR laser treatment shows the
tissue loss, (2) Silver staining reveals the localized nanoshells, (3) Hematoxylin/eosin staining shows tissue damage in
the similar area nanoshells localized, (4) MRTI calculations show similar dimensions of thermal damaged area [160].

Non-invasive SERS detection technique and photothermal effects of plasmonic
nanoprobes can be integrated to advance the remote control of the system. Gold nanorods (with
LSPR frequency of 790 nm) has been functionalized with IR-792 (λ max = 792 nm) Raman active
molecule along with PEG for SERRS detection (laser excitation at 785 nm) and photothermal
heating (laser excitation at 810 nm) (Figure 1.24) [164]. Bilateral human MDA-MB-435 tumor-
bearing mice model have been subcutaneously injected with SERRS-coded gold nanorods and
IR-792 Raman fingerprints have been detected at the tumor site. Upon IR irradiation at 810 nm,
gold nanorod injected tumors have been reached to ablative temperatures (>75 ˚C). In this study,
the SERS detection of plasmonic nanoprobes at the tumor site followed by photothermal therapy
has been successfully reported.
52

Figure 1.24 Raman active molecule modified gold nanorods (NRs) for NIR detection and photothermal therapy. (a)
Bilateral human MDA-MB-435 tumor bearing nu/nu mice were intratumorally injected with IR-792 modified gold
NRs. (b) In vivo Raman measurements reveals the Raman fingerprints of IR-792 modified gold NRs. (c) Infrared
thermographic map of mouse shows the significant temperature elevation with 810 nm laser irradiation in the tumor
region treated with IR-792-coded gold NRs [164].

Photoacoustic imaging and photothermal therapy can work in great harmony where
nanoprobes act as both contrast agents for imaging and thermal coupling agents for therapeutics.
It has been shown that PEG functionalized palladium/gold (Pd/Au) core/shell nanoplates have
been used for passive targeting of tumor tissue (4T1 murine breast cancer) and detected by
photoacoustic imaging (Figure 1.25) [165]. Pd/Au nanoplates with high blood circulation half-life
have shown dramatically higher accumulation on tumor tissue than any of the other organs. The
enhanced passive accumulation of tumor site has been attributed to the 2D geometrical structure
of the nanoprobes, besides the EPR effect. Photoacoustic images (700 nm excitation) have
revealed significantly increased nanoprobe accumulation in tumor tissue over time (0 h – 24 h)
(Figure 1.25a). Additionally, in this study, Pd/Au nanoplates have been used as computed
tomography (CT) contrast agents since gold nanoparticles possess high absorption at X-ray
wavelength range (Figure 1.25b) [166]. Finally, Pd/Au injected mice have shown dramatic
temperature elevation (>60 ˚C in 4 min) in tumor site upon 808 nm NIR irradiation compared to
uninjected mice (~3 ˚C) (Figure 1.25c). The recurrence of the tumors in Pd/Au injected and NIR
irradiated mice has not been observed, however untreated, only Pd/Au injected, or only NIR
irradiated control groups have shown tumor growth and ~20 days life span (Figure 1.25d). Here,
53

it has been clearly reported that Pd/Au nanoplates are versatile contrast agents for both
photoacoustic imaging and CT. Furthermore, due to the 2D geometry they reveal high tumor
accumulation compared to the nanostructures have been used in many previous studies. The high
accumulation in tumor site turns back as an advantage for photothermal therapy where the
thermal ablation threshold can be reached over a short irradiation time.

Figure 1.25 Core/Shell Pd/Au nanoplates as theranostic agents for in vivo photoacoustic imaging and photothermal
therapy. (a) In vivo photoacoustic imaging of Pd/Au nanoplate injected 4T1 murine breast cancer bearing mice.
Increased intensity of photoacoustic signal is shown over time. (b) In vitro CT signal increases as a function of
increased concentration of Pd/Au nanoplates. (3) In vivo cross-sectional CT image difference before and after injection
of Pd/Au nanoplates and (4) increased CT value after injection of Pd/Au nanoplates. (c) Infrared thermal images of
tumor bearing mice with Pd/Au treatment and control groups. Pd/Au treated mice shows an increased temperature in
tumor site upon 808 nm laser irradiation. (d) Digital photographs of mice with only Pd/Au injection and Pd/Au + laser
treatment [165].

1.3.5 Conclusion
In this chapter, three main application fields of plasmonic nanoprobes have been
summarized. First, we focused on SERS based biosensing application of nanoprobes. Highly
controlled nanostructure size and shape provides facile control of LSPR frequency which utilizes
in vitro detection experiments of different biochemical molecules. Furthermore, the high surface
area to volume ratio of nanostructures makes them suitable for multifunctional surface
modifications which bring targeted detections of molecules with high sensitivity and selectivity
54

compared to the conventional methods and high physiological stability of the nanoprobes.
Second, the contribution of nanoprobes in biomedical imaging has been explained, particularly in
SERS imaging and photoacoustic imaging techniques. There are many improvements on SERS
imaging in order to make the method faster for in vivo studies such as galvanometric mirrors and
angle-tunable filters. This non-invasive imaging technique can be equipped with such nanoprobes
that have LSPR frequency at NIR range where the highest depth of tissue penetration can be
achieved due to the reduced scattering and absorption of intrinsic tissue chromophores. It has
been shown in multiple studies that SERS imaging can be further advanced with multiplexing due
to the presence of many NIR Raman reporters and high detection sensitivity. Furthermore,
photoacoustic imaging is a rapidly emerging imaging method that utilizes thermal expansion due
to the optical excitation. High absorption cross section of plasmonic nanoparticles makes them
ideal contrast agents for photoacoustic imaging to improve non-invasive targeted detection of the
tumors. Not only SERS or photoacoustic imaging techniques, but also conventional methods,
such as MRI or CT, can use plasmonic nanoprobes as contrast agents. Third, therapeutic and
theranostic properties of SERS nanoprobes have a great potential use to eliminate the robust side
effects of chemotherapy and radiotherapy. Both photothermal therapy and photoacoustic imaging
utilize thermal emissions. These two systems can be combined in a single setup, even though,
while short pulse lasers are used for photoacoustic imaging, photothermal therapy requires
continuous wave lasers. However, the thermal expansion resulting in broadband pressure waves
does not cause significant heat production, and provides acoustic waves to detect the tumor tissue.
Multiple studies have shown successful tumor ablation using a combination of plasmonic
nanoprobes and NIR irradiation in a significantly short time. Most of them have revealed the
absence of tumor recurrence after the photothermal therapy which suggests high selectivity and
55

high absorption cross section of plasmonic nanoparticles make them promising new generation
therapeutic agents.
This section makes a connection between the engineered SERS nanoprobes in this thesis
(explained in CHAPTER 3) and their current and future application fields. In the next chapter, the
detailed characterization of fully folded fluorescent proteins and split fluorescent proteins, which
will be used as a multifunctional Raman reporter in CHAPTER 3, and quantitative analysis of
self-assembly reaction of split fluorescent proteins will be explained.
56

CHAPTER 2. CHARACTERIZATION OF NOVEL SPLIT
FLUORESCENT PROTEINS AND QUANTITAVIVE
ANALYSIS OF THEIR SELF-ASSEMBLY PROCESS
Introduction
Many biological applications rely on the use of split-fluorescent protein (split-FPs)
fragments to visualize protein-protein interactions, assemble nanomaterials, study the nanoscale
dynamics of individual biomolecules in cells, or identify synaptic interactions in vivo, among
many other uses (Figure 2.1) [167-169]. Surprisingly, little is known about the self-assembly
process of split-FPs and there is currently a lack of characterization concerning their biochemical
and biophysical properties including fragment binding rates, folding and chromophore maturation
rates and overall FP brightness following the interaction of complementary peptide fragments.
Here, we have developed a variety of split-FPs (sGFP, sYFP, sCFP) and their corresponding full-
length spectral variants (GFP, YFP, CFP), and we have comprehensively characterized their
photophysical and biochemical properties using fluorescence kinetic measurements, fluorescence
anisotropy, cell imaging, and single molecule biophysical techniques. We show that the in vitro
complementation of split-FP fragments follows a conformational selection mechanism whereby
the larger split-FP fragments exist in a monomer-dimer equilibrium and only monomers are
competent for binding the smaller peptide fragments. This bimolecular interaction involves a
slow and irreversible binding step of the complementary peptide followed by a fast maturation
step of the FP chromophore. Interestingly an excess of the small peptide fragment drives a slower
alternative complementation pathway involving a conformational change of the larger split-FP
fragment. Based on these in vitro characterizations, the brightest and the fastest maturing split-
FPs were expressed as fusion proteins in cells to assess their in vivo biochemical behaviors and
57

their respective performance for ensemble and single molecule cellular imaging. This study
resulted in development of novel split-GFP, split-YFP and split-CFP variants with improved
maturation kinetics, brightness, and photostability for Complementation Activated Light
Microscopy (CALM) imaging of cellular processes and for the controlled assembly of
nanomaterials.

Figure 2.1 Different application platforms of complementary split-GFP fragments. (a) The schematic of the GFP
(nano)polygon fabrication [167], (b) CALM imaging of M3 peptide-coated coverslips before and after addition of split-
GFP fragments, and individual-complemented split-GFP molecules (red circles) [168], (c) Identifying subcellular
locations of synapses by GFP reconstitution across synaptic partners (GRASP) [169].

58

Gene encoding for the green fluorescence protein (GFP) of the jellyfish Aequorea
victoria revolutionized fluorescent labeling applications [170] notably because GFP does not
require exogenous substrates or cofactors to produce bright fluorescent signals [171]. The gene
expression of a protein of interest fused to GFP allows the detection of various biomolecular
behaviors, such as protein-protein interactions and protein dynamics in vitro or in vivo [172].
Since the discovery of GFP 55 years ago [173], enormous progresses have been made toward
understanding its folding and fluorescence maturation kinetics. These advances have permitted
the development of hundreds of FP variants, including spectral variants, photoswitchable and
photoconvertible FPs as well as split-FP variants [172, 174], which have recently attracted a lot of
interests for single molecule (SM) microscopy imaging and nanomaterial assemblies [168]. Split-
FPs are on/off switchable and asymmetrically fragmented fluorescent probes engineered based on
a superfolder variant of GFP whereby the self-assembly of a large GFP 1-10 domain (referred to
as sGFP, amino acids 1-214) and a small GFP 11
th
β-strand peptide fragment (referred to as M3
peptide, amino acids 215-230) form a fully folded and mature GFP [175]. Protein of interest can
be expressed as fusions either to the sGFP or to the M3 peptide fragments, which are then
activated into a fluorescent and complete GFP upon introducing the missing complementary
fragment [168, 175]. Although it is very powerful, this split fragment self-assembly process and
the photophysical characteristics of split-FPs has not yet been clearly defined.

GFP structure and self-assembly of split-GFP
GFP forms a β-barrel [176] that encapsulates an internal p-
hydroxybenzylideneimidazolidinone chromophore which forms and matures after cyclization and
oxidation of three key amino acid residues in its sequence, namely Ser65-Tyr66-Gly67 residues
[98].
59

Figure 2.2 Ribbon diagram of the enhanced green fluorescence protein (eGFP) crystal structure (PDB entry: 2Y0G).
The α-helices are shown in purple, the β-strands are shown in yellow, the chromophore in the center α-helix is shown
as ball-and-stick model.

In wild-type GFP (wtGFP), the equilibrium of ionization states of the chromophore is
dictated by a complex hydrogen bonding network around the chromophore that gives rise to two
absorption maxima. One maximum absorbance peak at 395 nm is attributed to the neutral state of
the chromophore and is the dominant fraction of the population [172] for wtGFP at physiological
pHs. In this neutral state, Glu222 residue acts as a proton donor and forms a hydrogen bond with
Ser65 (Figure 2.3a). Glu222 also forms another hydrogen bond with Ser205 through a water
molecule [99]. Together with the major fraction of neutral chromophores observed in wtGFP, a
minor fraction of the population shows absorption at 475 nm, which has been attributed to the
anionic state of the chromophore [172]. In the anionic state, Glu222 is protonated by proton
abstracting through the hydrogen bond with Ser205 to form a hydrogen bond with Ser65, and
Glu222 transfers an electron to the chromophore. In order to stabilize the anionic state of the
chromophore, His148 residue forms a hydrogen bond with Tyr66. There is a loss of hydrogen
bond between Glu222 and Ser205 which destroys the hydrogen bonding network [99].
60

Because of the intimate contact between Glu222 and S65, introducing point mutations
such as S65T on this chromophore residue in wtGFP leads to alteration of the FP spectral features
and additionally makes wtGFP a more photostable and faster-folding protein [177]. S65T
mutation pushes the equilibrium towards the anionic state of the chromophore and results in a
more photostable chromophore with a 15 nm spectral shift in maximum absorbance peak (~490
nm) that can be assigned to the oxygen position in S65T structure (Figure 2.3b) [99, 171, 177].
Indeed, the methyl group in Thr65 causes a steric hindrance and results in a hydrogen bond
between Thr65 and Val61 rather than with Glu222. This interaction makes the anionic state of the
chromophore permanent by preventing ionization of Glu222.
Superfolder GFP variants of wtGFP, which include the S65T mutation together with
many additional mutations for improved folding (described below), also have a mature
chromophore locked in the anionic state [175, 178]. Importantly, when the C-terminal 11
th
β-
strand (M3-peptide) of the superfolder GFP β-barrel is removed, the resulting large sGFP
fragment cannot form a cyclized chromophore because this last β-strand encodes for Glu222,
whose interactions with Thr65 is required for chromophore cyclization and maturation [175].
Providing the missing M3 peptide to sGFP, however, results in complementation of a full-length
GFP, which can then mature and fluoresce. Although, the rates of protein folding and the
chromophore maturation steps have been well described for wtGFP [179, 180] and superfolder
GFP [178], the folding and maturation kinetics of the sGFP fragments upon interaction with M3
peptides have never been reported. Here, we have used the original sGFP (sGFPori) [175] to
elucidate the self-assembly kinetics between sGFP and M3 complementary fragments, and we
have further developed sGFP spectral and folding variants by point mutagenesis to improve
protein folding, assembly kinetics, chromophore maturation and spectral tunability of these
asymmetric split-FPs.
61

Figure 2.3 Illustration of chemical interactions between the GFP chromophore and the surrounding water molecules
and residues. (a) wtGFP and (b) S65T mutant [99].

Point mutations in split-GFP to generate split-GFP variants
The original sGFP (sGFPori) used in this study has folding reporter GFP mutations
(F64L, S65T, F99S, M153T, V163A), superfolder GFP mutations (S30R, Y145F, I171V,
A206V) that provides better solubility and increased complementation rate with the 11
th
β-strand
complementary fragment (M3 peptide), and additional mutations N39I, T105K, E111V, I128T,
K166T, I167V and S205T to increase brightness, solubility and fluorescence stability upon
complementation under physiological conditions [175] (Table 2.1). Using site directed
mutagenesis we made several point mutations in sGFPori and full-length GFPori (flGFPori) to
create sGFP variants. Various GFP variants that carry point mutations around the chromophore
have been reported to have either increased or decreased folding rate, to have improved
chromophore maturations rates, and to have improved brightness [181]. For instance, a GFP
variant called SuperGlo and carrying a Thr167 mutation has been shown to have improved
maturation rates [181]. Therefore, we tested a V167T mutation on sGFPori to create a sGFP1
variant. Combining V167T with mutation S72A and N149K such as those found in the Emerald-
62

GFP [111] was also shown to increase the chromophore maturation rate [181]. Therefore, we
tested a second, sGFP2 variant carrying both V167T and S72A mutations as well as a third
sGFP3 variant carrying all three V167T, S72A and N149K mutations with the aim of improving
the maturation rate of sGFPori.

Point mutations in split-GFP to generate split-YFP and split-CFP spectral variants
Based on the crystal structure of GFP, Thr203 amino acid residue locates in close
proximity to the chromophore [182]. T203Y mutation is essential to create split yellow
fluorescent protein (sYFP) variants. T203Y increases polarizability around the chromophore and
creates π-π interactions with the chromophore phenol ring (Tyr66) which lowers the excited state
energy and induces a red shift of the excitation and emission wavelengths. Also, Tyr66 based
chromophores require a symmetric chromophore cavity stabilized with hydrogen bond
interactions which are formed by Ser205 and Glu222 [183]. Based on these point mutations, we
developed a sYFP1 variant of sGFPori carrying T65G, T203Y, and T205S mutations (Table 2.1).
We also tested a T65L mutation together with the T203Y and T205S mutations to create a sYFP2
variant. Besides these two sYFP variants, we additionally developed a third sYFP3 mutant that
carries T203Y and T205A substitutions both of which have been reported to induce strong yellow
fluorescence of YFP [184].
Besides sGFP and sYFP mutants, we also developed two sCFP spectral variants (Table
2.1). In general, the chromophore of CFP is formed by a substitution of Tyr66 for Trp66, which
replace the phenol group in the chromophore with an indole [172]. Even though Y66W mutation
is sufficient to convert sGFP into a sCFP, the fluorescence emission is significantly lower [172].
However, it has been reported that additional mutations can restore the fluorescence intensity
[111]. The critical feature to obtain FPs with high quantum yield (QY) is to induce planar
63

conformations of the chromophore [185]. In sCFP variants, we therefore used an additional
H148D replacement that causes a strong repulsion against the indole ring of the chromophore
resulting in fully planar conformation (Figure 2.4) [186]. As H148D disrupts the van der Waals
interaction between the Y66W residue of the chromophore, the Ser205 residue becomes crucial to
stabilize the tryptophan-based chromophore through hydrogen bonding between the oxygen of
Ser205 and the polar nitrogen atom of Trp66. Furthermore, Thr203, Glu222 and Val150
neighboring residues stabilize the indole ring of the chromophore through van der Waals
interactions. On the other side of the indole ring, Val61, Thr62, Phe165, and Ile167 residues
provide van der Waals interactions to further support the chromophore. Beyond the importance of
stabilizing interactions, an additional mutation such as the E124V mutation has been reported as
one of the substitutions in GFP that allow faster folding rate than superfolder GFP [187]. Along
with the E124V mutation, D19E and D21E substitutions can also provide faster folding rates
compared to the sCFP carrying only Y66W and T205S mutations [184]. Therefore, we used
D19E, D21E, Y66W, H148D, and T205S mutations on both sCFP1 and sCFP2 variants. An
additional E124V mutation was used in the sCFP1 variant to improve the folding rate, and a
V167I mutation was applied in the sCFP2 variant to improve the QY.
64

Table 2.1 Fluorescence properties of full-length and split fluorescent proteins with mutations they carry.

Figure 2.4 Schematic of the interactions between the chromophore in CFP structure. The chromophore is stabilized by
van der Waals interactions (half circles) and hydrogen bonds (dashed line). H148D mutations plays a role in planarity
of the chromophore [186].

Exc λ
(nm)
Em λ
(nm)
εᶧ Φ Brightness τ1 (ns) A1% τ2 (ns) A2% Mutations
flGFPori 485 507 37700 0.66 24640 2.46 100 - 0
S30R, N39I, F64L, S65T, F99S, T105K, E111V, I128T,
Y145F, M153T, V163A, K166T, I167V, I171V, S205T,
flGFP1 485 507 33800 0.75 25270 2.5 100 - 0 ᵠ V167T
flGFP2 491 510 38200 0.77 29210 2.77 100 - 0 ᵠ V167T, S72A
flGFP3 482 508 39900 0.68 27290 2.67 100 - 0 ᵠ V167T, S72A, N149K
flYFP1 511 522 68300 0.61 41300 2.76 100 - 0 ᵠ T65G, T203Y, T205S
flYFP2 509 522 38200 0.54 20670 1.13 29 3.14 71 ᵠ T65L, T203Y, T205S
flYFP3 515 526 37100 0.51 18900 0.72 37 3.59 63 ᵠ T203Y, T205A
flCFP1 435 476 20400 0.41 8430 0.98 62 2.57 38 ᵠ D19E, D21E, Y66W, E124V, H148D, T205S
flCFP2 436 476 21300 0.42 8880 0.97 27 2.76 73 ᵠ D19E, D21E, Y66W, E124V, H148D, V167I, T205S
sGFPori 485 508 37600 0.59 22340 2.29 100 - 0
S30R, N39I, F64L, S65T, F99S, T105K, E111V, I128T,
Y145F, M153T, V163A, K166T, I167V, I171V, S205T,
sGFP1 485 508 33800 0.62 21010 2.34 100 - 0 ᵟ V167T
sGFP2 491 510 38200 0.67 25720 2.64 100 - 0 ᵟ V167T, S72A
sGFP3 480 508 39900 0.29 11600 2.5 100 - 0 ᵟ V167T, S72A, N149K
sYFP1 510 523 68300 0.44 30230 2.62 100 - 0 ᵟ T65G, T203Y, T205S
sYFP2 509 522 38200 0.09 3430 0.74 22 2.9 78 ᵟ T65L, T203Y, T205S
sYFP3 515 524 37100 0.35 12970 0.47 40 3.28 60 ᵟ T203Y, T205A
sCFP1 433 475 20400 0.18 3730 0.81 40 2.4 60 ᵟ D19E, D21E, Y66W, E124V, H148D, T205S
sCFP2 433 476 21300 0.23 4930 0.64 25 2.52 75 ᵟ D19E, D21E, Y66W, E124V, H148D, V167I, T205S
ᶧ Molar extinction coefficient (M¯¹cm¯¹)
Φ Quantum yield
ᵠ Additional mutations in eGFPori protein structure
ᵟ Additional mutations in sGFPori protein structure
τ1 Lifetime-1
τ2 Lifetime-2
A1 Fraction of lifetime-1
A2 Fraction of lifetime-2
65

Photophysical characteristics of full-length FP and split-FP variants
Optical spectra of full-length FP and split-FP variants
Split-FPs and their corresponding full-length FPs exhibited similar absorption and
emission spectra (Figure 2.5). sGFPori and sGFP1 (as well as flGFPori and flGFP1) had
absorption and emission spectra similar to those of Emerald-GFP [111]. In sGFP2 and flGFP2,
the S72A mutation caused a clear red-shift in absorption maxima and a slight red-shift in
emission maxima compared to the sGFPori (Table 2.1, Figure 2.5). The red-shift effect of S72A
was reversed by the addition of N149K mutation in sGFP3 and flGFP3.
Concerning our YFP variants, as stated before, a T203Y substitution is essential to induce
red-shifts of the absorption and emission spectra in GFP and create YFP. As seen in Figure 2.5, in
all our sYFP variants, the T203Y mutation effectively resulted in large spectral red-shifts of the
chromophore, which displayed absorption band maxima at around 510 nm (Table 2.1, Figure
2.5). Interestingly, an additional weaker absorption peak at 395 nm was observed for some of our
YFP variants. This weak 395 nm band was absent in the T65G sYFP1 and flYFP1 variant but
was present in sYFP2, flYFP2, sYFP3, and flYFP3 (Figure 2.5). Indeed, consistent with the key
role played by amino acid residue 65 for the dual peak excitation of wtGFP, the different T65G,
T65L, and S65T mutations in our three YFP variants, resulted in slightly different influences on
the appearance of the 395 nm band. For instance, in sYFP2 and flYFP2, where we use a T65L
mutation instead of T65G, the effect of Leu65 on the hydrogen bonding network is visible by the
presence of a small 395 nm absorption peak (Figure 2.5). Even though sYFP3 and flYFP3
variants carry S65T mutation which normally has a disruptive effect on 395 nm absorption peak,
sYFP3 has a more pronounced absorbance at 395 nm compared to sYFP1 and sYFP2 (Figure
2.5). This might be an effect due to the presence of the Ala205 residue since Ala is incapable of
66

hydrogen bond formation with Tyr66 to create a symmetric chromophore cavity, which resulted
in a partial spectral red-shift and the presence of 395 nm absorption peak.
In the case of the sCFP and flCFP variants, the Y66W mutation caused a blue-shift of the
absorption and emission spectra, turning GFP into a CFP spectral variant as expected. The V167I
difference between the sCFP1 and sCFP2 variants did not cause any significant spectral change
(Table 2.1, Figure 2.5).

Figure 2.5 Absorption and emission spectra of (a) split fluorescent proteins, and (b) full-length fluorescent proteins.

Fluorescence lifetime of full-length FP and split-FP variants
Beyond the optical spectra, fluorescent lifetime (τ) is a critical parameter to consider
when selecting a suitable fluorescent label for a particular application. Fluorescent lifetime is the
67

average time required by a population of electrons at an excited state to decay exponentially and
produce fluorescence as they lose energy. Absorption of photons by a fluorescent molecule
triggers a chain of photophysical events, such as internal conversion, fluorescence and inter-
system crossing [188]. Fluorescence lifetime is an intrinsic property of a fluorophore, hence it is
independent of the fluorescence intensity, the fluorophore concentration, or the method of
measurement. However, it highly depends on the fluorophore structure and external conditions,
such as temperature, polarity, pH, or refractive index of the surrounding media [189-191].
The chromophore of wtGFP has three electronic state forms; a neutral A-state, which is
excited at ~400 nm, a deprotonated B-state, which is excited at ~480 nm, and another
deprotonated intermediate state called I-state, which is excited at ~500 nm [192, 193]. Typically,
S65T-GFP variants display only the B and I electronic states, which have reported fluorescence
lifetimes of 2.8 ns and 3.3 ns, respectively [193]. Indeed, in S65T-GFP variants, the anionic
chromophore exists in either a benzoidal or a quinoidal form. In the benzoidal form, which is
attributed to the B-state, the majority of the negative charge on the chromophore phenol group is
stabilized by hydrogen bonding to Thr203. However, in the quinoidal form, which is attributed to
the I-state, this stabilization is not as efficient and changes in the chromophore structure result in
a prolonged fluorescence lifetime [194]. All our sGFP and flGFP variants possess the Thr203
residue and the resulting stabilization of the chromophore benzoidal form resulted in fluorescence
lifetime decays that were well described by a single exponential decay time constant (Table 2.1).
We found that flGFPori has a fluorescence lifetime of 2.46 ns, ~0.25 ns shorter than S65T-GFP
[191]. The V167T mutation of sGFP1 and flGFP1 did not significantly affect the lifetime of the
chromophore. However, sGFP2 and flGFP2 which carry the S72A mutation exhibited a slightly
higher fluorescence lifetime of 2.77 ns, indicating that S72A impacts the photophysics of the
chromophore, although the mechanism is not clear at the moment. On the other hand, the
68

additional N149K mutation in sGFP3 and flGFP3 slightly reversed the effect of S72A, with a
lifetime value decreasing to 2.67 ns. Interestingly, all the split-FPs exhibit ~200 psec shorter
lifetime than the full-length FPs (Figure 2.6). A possible reason for this phenomenon is that the
reformed chromophore of the split-FPs is more exposed to the surrounding medium due to the
lack of robust β-barrel protein structure, compared to the chromophore of the full-length FPs,
which is well protected in the center α-helix of the β-barrel structure.

Figure 2.6 Fluorescence lifetime (τ) comparison between full-length and split version of the same fluorescent proteins.
For YFP2, YFP3, CFP1 and CFP2 variants, only the longest and generally dominant lifetime values are reported here.
○ represents deprotonated B-state, ◊ represents neutral A-state and deprotonated B-state, * represents neutral A-state
and deprotonated I-state of the chromophore.

For our YFP variants, the presence of T203Y mutation is expected to induce significant
emission from the chromophore I-state with longer fluorescence lifetime values than for our GFP
variants [194]. Surprisingly, however, the fluorescence lifetime of sYFP1 and flYFP1 was best
described by a single exponential decay with a time constant of 2.76 ns, similar to the lifetime
expected for a chromophore in the B-state and in agreement with the previously reported lifetime
of a S65G/T203V YFP mutant [194]. This suggests that one of the two additional T65G or T205S
mutations stabilizes the hydrogen bonded benzoidal form of the chromophore. In comparison
both sYFP2/flYFP2 and sYFP3/flYFP3 displayed biexponential lifetime decays with values
69

around 3.3 ns and 0.5 ns. The longer 3.3 ns lifetime value is well within the range expected for
fluorescence emission from the I-state, while the shorter 0.5 ns lifetime corresponds to
fluorescence emission from the A-state. Indeed, similar lifetimes of 3.3 ns and ~0.8 ns were
previously assigned to the protonated I-state and the neutral A-state, respectively, for a
T203V/S205A YFP mutant [195]. The presence of an additional A-state short fluorescence
lifetime in sYFP2/flYFP2 and sYFP3/flYFP3 is fully consistent with the presence of the residual
absorption peak at 395 nm in these variants. Indeed, this absorption band, which corresponds to
the neutral A-state of the chromophore [192, 196] might have been excited by the two-photon 870
nm laser excitation used for the lifetime measurements assays. As a result, the fluorescence
emission detected at 515 nm likely stems from a combined contribution of both A- and I-states
for the YFP2 and YFP3 variants. We note that a simple change of G65 to L65 between the YFP1
and YFP2 resulted in a shift of the chromophore emission from a stabilized B-state to a less stable
combination of A- and I-state. This underlines the important role that G65 plays in stabilizing the
protonated benzoidal form of the chromophore when an additional T203Y mutation is present,
such as in YFP1. This enhanced chromophore stabilization and its likely increased planarity are,
in fact, in full agreement with the larger quantum yield and the larger extinction coefficient of the
YFP1 variant compared to the two other YFPs we have studied. As for the GFP variants, sYFPs
displayed ~200 ps shorter fluorescence lifetime decays than flYFPs.
For our CFP variants, we observed biexponential lifetime decays, in agreement with the
presence of the two chromophore lifetimes previously reported for ECFP [197]. As reported by
other, the Y66W substitution, which is present in both our CFP variants, is expected to induce
emission from the A-state of the chromophore [192, 197, 198] with a reported A-state lifetime of
2.52 ns for ECFP [199, 200]. Consistent with these observations, the longest lifetime values of
flCFP1 and flCFP2 were 2.57 ns and 2.76 ns respectively, indicating significant emission from
70

the chromophore A-state. flCFP1 and flCFP2 also had much shorter life time at 0.98 ns and 0.97
ns respectively, again in good agreement with the shorter lifetime of ECFP which was reported at
~0.6 ns [197]. This shorter lifetime is associated with the minor B-state conformation of the
chromophore where Tyr145 residue is moved outward from the chromophore and H148 residue is
turned towards the interior [197]. In the longer lifetime A-state, however Tyr145 is turned
inwards towards the chromophore while, H148 is moved outward. Even though it has been shown
that H148D-ECFP mutant exhibits longer fluorescence lifetime [199, 200], additional mutations
in our CFP variants appear to suppress this H148D contribution. As for the GFP and YFP
variants, sCFPs displayed ~200 ps shorter fluorescence lifetime decays than flCFPs.

Fluorescence brightness of full-length FP and split-FP variants
Another important characteristic of a fluorophore is brightness, which represents the
number of emitted photons for a given number of absorbed photons. The brightness is the product
of the molar extinction coefficient (ε) and the quantum yield (Φ) of the fluorophore. In order to
calculate molar extinction coefficient of each FP, Beer-Lamber law was used:

𝐴 = 𝜀𝐶𝑙 2.1

where A is the absorbance, C is the concentration of the fluorophore, and l is the length of the
light path through the fluorophore sample. Relative quantum yield of each FP was calculated
using a fluorescein standard with 0.93 quantum yield [201, 202]. The optimal excitation
wavelength was defined as shown in Figure 2.7, in order to avoid the spectral overlap in the red
tail between the scattered excitation wavelength and the emission.

71

Figure 2.7 Overlapped absorption spectra of the standard (Ast) and the sample (Ax). The red circle shows the optimal
excitation wavelength that has been used for quantum yield calculations [201].

First, the absorption factors of the fluorescein standard (f st) and the FP samples (f x) were
calculated using measured absorbance (A) at the excitation wavelength (λ ex is 460 nm for GFPs,
440 nm for CFPs, 475 nm for YFPs) using:

𝑓 = 1 − 10
−𝐴 (𝜆 𝑒𝑥
)
2.2

The absorption factors can also be calculated with improved accuracy using a band-pass
component, Δλ ex, for the excitation wavelengths. Here we have used ±1 nm band-pass:

𝑓 = ∫ (1 − 10
−𝐴 (𝜆 𝑒𝑥
)
)𝑑 𝜆 𝑒𝑥
𝜆 𝑒𝑥
+1/2Δ𝜆 𝑒𝑥
𝜆 𝑒𝑥
−1/2Δ𝜆 𝑒𝑥
2.3

Second, the relative integral photon fluxes emitted from the fluorescein standard (F x) and the FP
samples (F st) were calculated based on the spectrally corrected and blank-corrected spectrum of
each sample (I Cʹ) using:

𝐹 = ∫ 𝐼 𝐶 ⋅ 𝜆 𝑒𝑚
𝑑 𝜆 𝑒𝑚
𝜆 𝑒𝑚

2.4

72

Third, the fluorescence quantum yield was calculated using:

Φ
𝑓 ,𝑥 = Φ
𝑓 ,𝑠𝑡
⋅
𝐹 𝑥 𝐹 𝑠𝑡
⋅
𝑓 𝑠𝑡
𝑓 𝑥 ⋅
𝑛 𝑥 2
(𝜆 𝑒𝑚
)
𝑛 𝑠𝑡
2
(𝜆 𝑒𝑚
)
2.5

where F is the emitted relative integral photon flux, f is the absorption factor, n is the refractive
index of the solvent, Φ f,x is the quantum yield of the sample, Φ f,st is the quantum yield of the
standard. The calculated ε and Φ values of each protein were used to reckon the brightness (Table
2.1). From these measurements, we have found that both split and full-length version of the same
FP variants have the same extinction coefficient values (Table 2.1).

Figure 2.8 Brightness comparison between all fluorescent proteins.

We found that the extinction coefficient of flGFPori (37700 M
-1
cm
-1
) is similar to that
previously reported for S65T-GFP (39200 M
-1
cm
-1
) [101]. The quantum yield of flGFPori (0.66)
is also higher than the EGFP (0.60) which carries the same F64L/S65T mutations [203]. Further
improvement on quantum yield was achieved with flGFP2 (0.77, +17%), which, in fact, has a
higher quantum than the protein Emerald-GFP (0.68, S65T/S72A/N149K/M153T/I167T) [203].
When flGFPori and flGFP1 variants were compared, the V167T substitution seemed to increase
73

the quantum yield but also slightly decreased the extinction coefficient resulting, nonetheless, in
an improved brightness of flGFP1 compared to flGFPori (Figure 2.8). The additional S72A
mutation in the flGFP2 variant increased the extinction coefficient, which, together with the
improved quantum yield, resulted in flGFP2 (V167T/S72A) having the highest brightness among
all flGFP variants (Figure 2.8). The N149K substitution in flGFP3 further increased the ε,
however it slightly decreased the Φ, and the overall brightness of flGFP3 was decreased
compared to the flGFP2 variant (Figure 2.8). Overall, flGFP2 was the brightest of all the flGFP
variants, followed by flGFP3, flGFP1, and the least bright was flGFPori (Figure 2.8). Similar
optical properties were observed for the split-FP variants. sGFP2 had the highest quantum yield
among all sGFP variants with a value of 0.67, which corresponds to a 14% improvement
compared to sGFPori. Surprisingly, however, the N149K mutation in sGFP3 had a dramatic
effect on the Φ value, which dropped to 0.29 compared to 0.68 in flGFP3. Consequently, sGFP3
was found to be the least bright among all sGFP variants. Overall, sGFP2 showed the highest
brightness among the other sGFPs, followed by sGFPori, sGFP1, and sGFP3 (Figure 2.8).
For the YFP variant, we found that flYFP1 had a quantum yield of 0.61 consistent with
previously reported values for YFP (0.61) [203, 204]. sYFP1 and flYFP1 mutants have the
highest ε value compared to all proteins, and ended up being the brightest FPs (Figure 2.8),
despite the fact that they do not have high Φ values. The massive increase in ε of YFP1 variant
was attributed to the T65G substitution, because T65L-YFP2 mutant had lower ε. The presence of
T65L/T205S (YFP2) or S65T/T205A (YFP3) did not result in significant difference on ε, Φ, and
brightness for flYFP2 and flYFP3. However, the T65L mutation dramatically decreased the Φ of
sYFP2, which had the lowest brightness among all the FPs.
Our CFP variants showed the lowest extinction coefficients and quantum yields among
the different FP variants studied. flCFP1 and flCFP2 had quantum yield values of 0.41 and 0.42,
74

respectively. These values are similar to that of a Y66W-CFP mutant called W1B, which has
mutations similar to that of our CFP variants and a quantum yield of 0.4 [203]. However, the
extinction coefficients of our CFP variants (flCFP1: 20400 M
-1
cm
-1
, flCFP2: 21300 M
-1
cm
-1
) are
significantly lower than this W1B mutant (32500 M
-1
cm
-1
) [203]. The addition of V167I mutation
on CFP2 slightly increased the extinction coefficient, therefore resulting CFP2 being brighter
than CFP1.

Fluorescence photostability of full-length FPs and split-FPs
The photobleaching characteristics of FPs also provide crucial insights into their
photophysical properties and are important parameters to consider when choosing appropriate
FPs for applications such as single molecule imaging, fluorescence resonance energy transfer
(FRET) or fluorescence recovery after photobleaching (FRAP). Photobleaching reactions occur in
two phases: (i) a photoconvertible dark state phase where a rapid but reversible decrease in
fluorescence intensity takes place, and (ii) an irreversible photobleaching phase where a slow and
continuous decrease in fluorescence is observed upon photodestruction of the chromophore
(Figure 2.9a) [205]. Both phases can significantly influence fluorescence measurements. For
instance, rapid irreversible photobleaching of fluorophores can prevent single molecule tracking
of individual biomolecules over long distances in cells. In addition, the spontaneous or light-
induced recovery of fluorescence due to the reversible photoconvertible dark state generally
appears as blinking events during single molecule imaging [205, 206], which further impact
tracking. In FRAP experiments, where fluorescent recovery in irreversibly photobleached areas
takes place due to the lateral diffusion and exchanges of bleached and unbleached fluorescently
labeled proteins, the reversible photoconvertible dark state might distort FRAP analyses [205]. In
FRET experiments, when the photobleaching rate of the acceptor is faster than the donor
75

photobleaching kinetics, FRET intensity is reduced due to lower availability of acceptor
molecules in ground state. Because understanding the photostability of FPs is important for these
analyses, we have conducted photobleaching kinetic assays where a constant excitation of the
complemented sGFP variants was applied during 40 minutes long acquisitions of their emission
(Figure 2.9b). The photobleaching process was fitted by the analytical solution to the following
differential equations that describe the photobleaching kinetics in Figure 2.9a:

𝑑 [𝐹𝑃
𝑛𝑎𝑡 ] 𝑑𝑡 ⁄ = −(𝑘 1
+ 𝑘 3
)[𝐹𝑃
𝑛𝑎𝑡 ] + 𝑘 2
[𝐹𝑃
𝑟𝑏𝑙𝑒 ]
𝑑 [𝐹𝑃
𝑟𝑏𝑙𝑒 ] 𝑑𝑡 ⁄ = −(𝑘 2
)[𝐹𝑃
𝑟𝑏𝑙𝑒 ] + 𝑘 1
[𝐹𝑃
𝑛𝑎𝑡 ]
2.6

where [FP nat] represents the concentration of FPs in their native state, [FP rble] represents the
concentration of reversibly bleached FPs and is assumed to be 0 at initial conditions, k 1 represents
the forward rate constant toward the photoconvertible dark state, k 2 represents the backward rate
constant from the photoconvertible dark state toward the native FP state, and k 3 represents the rate
constant of irreversible photobleaching. We also assume that [FP ible], the concentration of
irreversibly bleached FPs, was 0 at initial conditions [205].
As can be seen in Figure 2.9c, sGFPori shows the fastest k 1 rate towards the
photoconvertible dark state and sGFP1 has the slowest rates for both photobleaching steps
(Figure 2.9c). Indeed, the V167T mutation in sGFP1 resulted in ~46 % slower rate of
photoconvertible dark state compared to sGFPori, while their backward rates from the dark state
(k 2) are nearly identical. The addition of S72A (sGFP2) or S72A/N149K combination (sGFP3)
increase the k 1 rate toward the dark state compared to sGFP1, but both mutations still show ~24%
slower photoconvertible dark state rate than sGFPori (Figure 2.9c). Thus, sGFPori is more rapidly
converted to a dark reversible state than sGFP2, sGFP3 or sGFP1. On the other hand, return from
the dark state (k 2) is the slowest for sGFPori, followed by sGFP1, sGFP2 and sGFP3, which has
76

the fastest k 2 rate constant. These data indicate that sGFPori is prone to rapidly enter and spend
more time in the reversible dark state than all other sGFPs. Concerning the irreversible
photobleaching rates (k 3), the V167I mutation in sGFP1 slows down irreversible photobleaching
by 15 % compared to sGFPori. However, the addition of S72A mutation in sGFP2 canceled this
effect, which is partially recovered in by the S72A/N149K mutations in sGFP3. Overall,
complemented sGFPori shows the highest probability to be trapped in the photoconvertible dark
state and photobleaches faster than complemented sGFP1 or sGFP3. Complemented sGFP2, is
less prone to enter and stay in the photoconvertible dark state compared to sGFPori but has
similar irreversible photobleaching rate. Both sGFP1 and sGFP3 photobleach slower than
sGFPori or sGFP2 and are less prone to be trapped in the photoconvertible dark state under
continuous excitation.

77

Figure 2.9 Photobleaching kinetic of sGFP variants. (a) The model of irreversible photobleaching and photoconvertible
dark state reactions [205]. (b) Normalized kinetic measurement for photobleaching and photoconvertible dark state
reactions of sGFPori (black), sGFP1 (red), sGFP2 (blue), and sGFP3 (green) variants. (c) Comparison of faster
photoconvertable dark state rate constants (left panel), reverse photoconvertable dark state rate constants (center panel)
and slower irreversible photobleaching rate constants (right panel).

Folding and maturation rates of full-length FP variants
In addition to understanding the photophysical properties of our FP variants, it is also important to obtain detailed
information on their folding and chromophore maturation processes, which are determinant factors to understand how
full-length FPs can efficiently be re-assembled from separated split-FP fragments. Heim et al. have proposed a three-
step mechanism for wtGFP folding and chromophore maturation (
Figure 2.10) [172] which was later used for a detailed analysis of the rates of each step in
S65T-GFP mutants [180].

Figure 2.10 Three-step mechanism of GFP chromophore formation model. Folding step triggers the tripeptide
chromophore cyclization which involves nucleophilic attack of the amino group of Gly67 on the carbonyl group of
Thr65. Cyclization step is followed by an oxidation forms the mature GFP chromophore [172, 180].

The folding rate of full-length and mature FPs can be determined by monitoring the
fluorescence recovery of the chromophore after FP denaturation at 95 ˚C for 10 min in 8 M urea,
followed by dilution-induced refolding in denaturant-free buffers [179, 207]. This treatment,
which induces a loss of the GFP native structure, exposes the mature and chemically intact
chromophore to the surrounding environment and results in fluorescence emission quenching
[180, 208]. When the denatured FP regains its tertiary structure, the chromophore becomes
protected against the environment and emits fluorescence. The kinetic of fluorescence recovery is
therefore directly related to the rate of refolding. We have monitored the renaturation of our full-
length FP variants by acquiring the fluorescent emission (at maximum λ em) for 120 min from 100-
fold dilutions of the denatured full-length FPs in denaturant-free buffer. The folding mechanism
78

kinetic was fitted by a triexponential growth function (Equation 2.7) which defines three
independent first-order kinetic constants:

𝑦 = 𝐴 1
(1 − 𝑒 (−𝑘 1
𝑥 )
) + 𝐴 2
(1 − 𝑒 (−𝑘 2
𝑥 )
) + 𝐴 3
(1 − 𝑒 (−𝑘 3
𝑥 )
) 2.7

Each full-length FP shows a fast (k 1), a medium (k 2), and a very slow (k 3) first-order rate
constants. Both the fast (k 1) and the medium (k 2) rate constants of flGFPori are similar to those
observed previously for urea-unfolded S65T-GFP (k 1flGFPori of 1.556 min
-1
vs. k 1S65TGFP of 1.470
min
-1
and k 2flGFPori of 0.147 min
-1
vs. k 2S65TGFP of 0.146 min
-1
) [180] and were assigned to two
parallel folding pathways involving differently isomerized proline residues. Indeed, the profusion
of proline residues in FPs can impact their folding rates, with a fast folding rate constant (k 1fold)
for FP molecules which already have properly isomerized proline residues and a slower folding
rate constant (k 2fold) for FP molecules with improper conformation of proline residues that
undergo slow isomerization at the start of the renaturation process [180]. The very slow rate
constant (k 3) (its amplitude was typical 10% of the total kinetic recovery) was assigned to the
chromophore maturation constant (k mat), as it systematically matched the maturation rate
constants determined independently in reduced chromophore maturation kinetics assays (Figure
2.11a). This k 3 rate constant suggests the presence of maturing FPs in our refolding kinetics and
indicates that a small amount of none-matured FPs might have been present in the samples or that
urea denaturation and boiling led, to a small extent, to chromophore denaturation.
We found that the flGFPori showed folding rate constants (k 1fold: 1.556 min
-1
and k 2fold:
0.147 min
-1
) remarkably similar to those reported for S65T-GFP by Reid and Flynn [180]. As
seen in Figure 2.11b, the fast folding rate constant of flGFPori was improved by the addition of
V167T (flGFP1) or V167T/S72A (flGFP2) mutations. However, the combination of
V167T/S72A/N149K (flGFP3) decreased the initial improvement on k 1fold. Furthermore,
79

improved k 2fold rate constants were observed for the three flGFP1, flGFP2 and flGFP3 variants.
For the YFP variants, the fastest k 1fold rate constants were 0.551 min
-1
for flYFP1, 0.429 min
-1
for
flYFP2 and 0.363 min
-1
for flYFP3, which is higher than the previously reported folding rate
constant of EYFP (0.24 min
-1
) [209]. This large improvement of the folding rate is likely the
result of the many folding reporter and superfolder GFP mutations that were kept in all three
proteins, although the additional T205S substitution in flYFP1 and flYFP2 might participate to
their further improved folding rate compared to flYFP3. For the CFP variant, the fast folding rate
constant of flCFP1 (0.464 min
-1
) is consistent with the previously reported k 1fold of ECFP (0.66
min
-1
) [209]. The additional V167I mutation in flCFP2 dramatically improved both this fast
(1.955 min
-1
) and the slower (0.195 min
-1
) folding rates, making it significantly better folder than
ECFP.
In addition to the FP folding rates, the chromophore maturation rates can be determined
considering that maturation involves a two-step mechanism with a rapid cyclization step of the
Thr65-Tyr66-Gly67 residues followed by a rate limiting oxidation step [172, 180]. Using a strong
reducing agent, such as 5 mM dithionite, on the chromophore during FP denaturation provides a
simple means to monitor this oxidation step after dilution of the denatured and reduced proteins
in denaturant-free and dithionite-free buffers [180]. With this approach, we evaluated the
chromophore maturation kinetics of each full-length FP variants by fitting the fluorescence
recovery of the reduced chromophore using a single exponential growth function (Equation 2.8)
to extract the single phase and irreversible first-order maturation kinetic:

𝑦 = 𝐴 (1 − 𝑒 (−𝑘 𝑚𝑎𝑡
𝑥 )
) 2.8
As mentioned above, we found that the maturation rate constants (k mat) are similar to the very
slow rate constants (k 3) determined in the folding kinetic experiments. In Figure 2.11a, we have
displayed the folding and maturation kinetics of flGFP2 as a representative mechanism of fast
80

folding steps followed by a slow maturation step. As seen in Figure 2.11b, the mutation we have
implemented resulted in clearly improved maturation rates of flGFPori. For instance, while
flGFPori and flGFP1 have the same maturation rate constants (0.018 min
-1
or 55 min), both
flGFP2 (0.022min
-1
or 45 min) and flGFP3 (0.025min
-1
or 40 min) mature significantly faster.
Thus, while the V167T mutation had no clear effects on the chromophore maturation rate, both
S72A and N149K mutation provided improved maturation as expected [181]. These maturation
rates, however, are slightly slower than previously reported for similar GFPs [181]. For example,
the maturation rate of Emerald-GFP has been reported to be 0.084 min
-1
which is ~4 fold faster
than maturation rate of flGFP2. This might be an effect of the additional superfolder mutations in
our FPs. On the other hand, Reid and Flynn have reported a maturation rate constant of S65T-
GFP of 0.009 min
-1
that is ~2 times slower than flGFPori [180]. For the YFP variants, flYFP1
(0.013 min
-1
) and flYFP2 (0.009 min
-1
) were significantly slower maturing proteins than the GFP
variants, but flYFP3 appeared to have a very fast maturation with a rate constant of 0.096 min
-1
(10 min). This is a clear improvement compared to other YFP such as EYFP (0.043 min
-1
) or
Venus (0.025 min
-1
) [181]. For the CFP variants, flCFP1 (0.011 min
-1
) and flCFP2 (0.012 min
-1
)
had maturation rate constants similar to that previously reported for ECFP (0.0096 min
-1
) [209].
The V167I mutation in flCFP2 did not have any effect on maturation rate constant (Figure 2.11b).
81

Figure 2.11 Folding and maturation kinetic model of full-length FPs. Unfolded protein with oxidized chromophore can
be renatured to monitor folding kinetic mechanism. Unfolded protein with reduced chromophore can be renatured and
reoxidized upon dilution from denaturant. (a) Folding (red) and maturation (black) kinetics of flGFP2 were fitted by
triexponential and single exponential (green lines) growth functions, respectively. (b) Comparison of kfold1, kfold2, kmat,
and renaturation efficiencies of all the full-length FPs.

Self-assembly process of split-FPs
Having characterized the folding and maturation process in full-length FPs, we then
studied the self-assembly kinetics of complementary split-FP fragments and the maturation of the
FP chromophore upon their interaction. As a starting model, we assumed a partially folded
structure of the large split-FP fragment, an irreversible binding of the complementary M3 peptide
to the large fragment [168, 175] and an irreversible and independent maturation step of the
chromophore upon formation of the split-FP/M3 peptide complex (Figure 2.12). To characterize
this modeled self-assembly process, we conducted two different sets of pseudo-first order
reactions. In the first set of reactions, a constant M3 peptide concentration was reacted with
increasing concentration of the split-FP large fragment (sFP titration). In the second set of
82

reactions, a constant concentration of the split-FP large fragment was reacted with increasing
concentrations of M3 peptide (M3 peptide titration). Although similar self-assembly behaviors
were expected for both set of reactions under the simple model of Figure 2.12, quite different
complementation kinetic processes were observed for sFP titration and M3 peptide titration
(Figure 2.12a, b). Indeed, while the complementation kinetic appears to saturate within about 200
min at high sFP concentrations for the sFP titration (Figure 2.12a), the same kinetic is
significantly slower and does not fully reach saturation at long incubation time despite high M3
peptide concentrations in the M3 peptide titration (Figure 2.12b). In addition, while kinetic curves
for the sFP titration are well fit with a two-exponential growth model, those of the M3 peptide
titration require a triexponent fit.

Figure 2.12 Simple model of chromophore formation mechanism for split-FP fragments. (a) The first condition of
chromophore formation kinetic with sGFP2 titration and 100 nM M3 peptide fragment. (b) The second condition of
chromophore formation kinetic with 0.5 μM sGFP2 and M3 peptide titration.

Part of the difference between the two types of kinetics is due to the fact that the large
sFP fragment expressed in vitro exist in a concentration-dependent dimer-monomer equilibrium.
83

Indeed, using size exclusion chromatography (SEC), we have found that split-FPs at high
concentrations are mostly dimeric as previously reported by Cabantous et al. [175], but that
monomeric split-FPs become the dominant fraction at lower protein concentrations. To
characterize the concentration-dependent split-FP dimer:monomer equilibrium, we performed
fluorescence anisotropy experiment on sGFP2 and sGFPori at different protein concentrations.
Dimeric protein structures exhibit higher anisotropy values due to their slower diffusional
mobility than monomeric proteins. For these fluorescence anisotropy measurements, the split-
FPs, which were all expressed with an N-terminal tetracysteine (Cys 4) group, were labeled with
ReAsH (Invitrogen, T34562), a fluorescent dye particularly well-suited for anisotropy
measurements because of its rigid attachment to genetically encoded Cys 4 groups [210].
Fluorescence anisotropy values were calculated as follows:

𝐴 =
𝐼 ‖
− 𝐺𝐼
⊥
𝐼 ‖
+ 2𝐺𝐼
⊥
2.9

where A is the fluorescent anisotropy value, I ‖ is the parallel polarization intensity, I ⊥ is the
perpendicular polarization intensity, and G is the sensitivity correction factor of the instrument for
the two detection modes [211]. 1:2 molar ratio of sFP to ReAsH dye (40 μM:80 μM) was
prepared for 1.5 hour at room temperature and excess ReAsH were removed using Sephadex gel
filtration G-10 spin column (Harvard Apparatus). Endpoint equilibrium fluorescent anisotropy
measurements were acquired from 6 different concentrations of ReAsH-sFP complex using
540/25 excitation and 620/40 emission filters. As shown in Figure 2.13, as the concentration of
sGFP2 decreases, the fluorescent anisotropy of ReAsH-sGFP2 decreases which indicates that
more sGFP2 monomers are present at equilibrium and that split-FPs expressed in vitro effectively
exist in concentration-dependent monomer-dimer equilibrium. By fitting the equilibrium
84

anisotropy curve as a function of total protein concentrations with an equation that takes into
account the expected contribution of dimer anisotropy (rd), the contribution of monomer
anisotropy (rm) and their respective amount at each total protein concentrations tested (see
inserted equation in Figure 2.13), we defined the dimer equilibrium constant (Keq) for both
sGFP2 and sGFPori. For sGFP2 Keq of dimer formation was 0.7±0.6 μM
-1
and that of sGFPori
was quite similar at 0.32±0.2 μM
-1
.

Figure 2.13 Fluorescence anisotropy with respect to the concentration of ReAsH labeled sGFP2. Endpoint fluorescence
anisotropy was acquired from 0.5 μM, 1.25 μM, 2.5 μM, 5 μM, 10 μM, 20 μM ReAsH-sGFP2 complex. Increased
concentration of ReAsH-sGFP2 causes increased anisotropy values as an indication of forming more dimeric
structures. There is a concentration dependent monomer-dimer equilibrium for in vitro sFP samples.

Using the now defined Keq constants of dimer formation, the actual concentration of
monomers was evaluated in our sFP titration kinetic data (Figure 2.12a), and kinetic curves were
fitted with a simple biexponential growth function:

𝑦 = 𝐴 1
(1 − 𝑒 (−𝑘 𝑜𝑏𝑠 1
𝑥 )
) + 𝐴 2
(1 − 𝑒 (−𝑘 𝑜𝑏𝑠 2
𝑥 )
) 2.10

85

where k obs1 and k obs2 represent two independent observed rates corresponding to fragment binding
and chromophore maturation steps respectively, as implied by the simple model of Figure 2.12.
As shown in the case of sGFP2, k obs1 is linearly dependent on the sGFP2 monomer
concentrations, while k obs2 is constant and independent of monomer concentrations (Figure
2.14a). Fitting the distribution of k obs1 with a simple linear function resulted in the determination
of a k on binding rate constant of 0.0032 µM
-1
min
-1
with a y intercept at 0, fully consistent with
the irreversible binding between the large sGFP2 fragment and a M3 peptide. Fitting the
distribution of k obs2 with a constant resulted in the determination of the sGFP2 maturation rate
constant k mat of 0.025 min
-1
, in very good agreement with the maturation rate of full-length GFP2
determined independently (0.022 min
-1
, Figure 2.11b). This indicates that the maturation rates of
split- and full-length FPs are very similar. The calculated k on values (5.3x10
1
M
-1
sec
-1
for sGFP2)
are significantly smaller than the rates expected for diffusion-limited reactions, which, for typical
protein/protein associations, lie within 10
4
-10
6
M
-1
sec
-1
[212]. This 3-5 order of magnitude
difference suggests that k on represents a slow steric fit process between the complementary split-
FP fragments that takes place after diffusion and collision, both of which are nonetheless required
for the two fragments to interact. Thus, the complementation reaction appears to be primarily
limited by a slow steric fit step rather than by the time of the diffusion in solution. Ways to
further assess the importance of both the diffusion and steric fit processes in the complementation
kinetic would involve: (i) performing complementation assays in viscous solutions (e.g. 50%
glycerol) to assess the impact of diffusion, or (ii) performing complementation assays with
slightly modified peptide sequences for the M3 peptide to test the impact of steric fit. Under
similar sGFPori titration conditions, a k on of 0.0042 µM
-1
min
-1
was observed together with a
constant k mat of 0.016 min
-1
, again in good agreement with the maturation rate of full-length
GFPori (0.018 min
-1
, Figure 2.11b). This indicates that while the on-binding rate of sGFPori
86

fragments is slightly faster than that of sGFP2, the chromophore maturation rate is slower, by
about 20 minutes. Notice that for both sGFP2 and sGFPori, there is no influence of the
dimer:monomer equilibrium on the complementation kinetics during sFP titrations because there
is always a sufficient excess of sFP monomers, so that the reaction rates are not impacted by the
forward (k f) or the backward (k r) rate constant of dimer:monomer exchange. As such, the
complementation kinetic of our asymmetric split-FP system appears to follow a conformational
selection process. Indeed, despite the presence of dimers and monomers of the larger split-FP
fragments in solution, only the monomers are competent for a slow and irreversible binding to the
small M3 peptide fragment. This binding is followed by an irreversible and fast maturation step
of the FP chromophore (Figure 2.14). Additional measurements of k on and k mat were performed
for some of the other split-FP fragments variants as presented in Figure 2.14b, which shows that
the maturation rate of the different split-FP variant is generally faster than that of sGFPori, as
initially observed for full-length FPs.

87

Figure 2.14 Proposed model of split-FP chromophore formation mechanism under the condition of sFP titration. (a)
sGFP2 titration kinetics were fitted by biexponential growth function, and pulled out kobs1 and kobs2 were plotted as a
function of calculated monomeric sGFP2 concentration based on the monomer-dimer equilibrium ratio. kobs1 and kobs2
were assigned as kmat and kon, respectively. (b) Comparison of kmat and kon for all split-FPs. NYD: Not yet determined.

We then focus on our other M3 peptide titration kinetic measurements (Figure 2.12b)
which were performed with a total sGFP2 concentration of 0.5 μM (equivalent to 0.34 μM of
monomeric sGFP2). In these assays, the kinetic curves required fitting with the following
triexponential function:

𝑦 = 𝐴 1
(1 − 𝑒 (−𝑘 1
𝑥 )
) + 𝐴 2
(1 − 𝑒 (−𝑘 2
𝑥 )
) + 𝐴 3
(1 − 𝑒 (−𝑘 3
𝑥 )
) 2.11

where k obs1, k obs2 and k obs3 represent three independent observed rates. When plotted as a function
of M3 peptide concentration, the distribution of one of the observed rate, k obs1, shows a linear
response with peptide concentrations, with a slope of 0.0038 μM
-1
min
-1
very similar to the
irreversible k on rate constant determined during sFP titrations (Figure 2.15a). Interestingly, at the
lowest peptide concentrations tested, k obs1 was relatively constant and independent of peptide
concentration, hovering around a value of 0.021-0.022 min
-1
, which corresponds to the expected
maturation rate of sGFP2. This constant value of k obs1 at low M3 peptide concentrations was thus
assigned to k mat. This distribution of k obs1 can be interpreted as follows: At very low M3 peptide
concentrations (e.g. 0-0.5 μM), our pseudo-first order condition breaks down because there is not
a sufficient excess of M3 peptides over sGFP2 monomers (0.34 μM). In this case, k obs1 likely
reflects the fastest possible rate of fluorescence appearance independent of peptide concentration,
which, as we have discussed above, corresponds to k mat. As the concentration of M3 peptide
increases we recover our pseudo-first order condition and k obs1 is the compounded value of k mat
plus k on*[M3 peptide], which explains why k obs1 becomes linearly dependent on the concentration
of M3 peptides.
88

The presence of the two additional observed rates k obs2 and k obs3, indicates that there are
additional and unidentified steps in the fragment association and the chromophore formation
mechanisms that we did not observe during the sFP titration. When k obs2 and k obs3 are determined
at each kinetic measurement and plotted with respect to the increasing concentration of M3
peptide, they appear to both level off at final k obs2 values of 0.01 min
-1
and final k obs3 values of
0.0022 min
-1
and this despite the use of large M3 peptide concentrations (Figure 2.15b). This
suggests that, in addition to the k on binding and k mat rate constants, there are two additional and
rate-limiting rate constants affecting the complementation kinetics performed by M3 peptide
titration. Considering the dimer:monomer equilibrium of sGFP2, it is likely that one of these two
additional constants represents the forward constant of dimer dissociation k f, which could indeed
be rate-liming. To understand the origin of both these slow k obs2 and k obs3 rates, we performed
additional kinetic fluorescence anisotropy measurements on sGFP2. A fixed concentration of
sGFP2 at 0.5 μM was incubated with increasing concentrations of M3 peptide and the anisotropy
of each sample was measured over periods of 13-14 hours (Figure 2.15c). Changes in anisotropy
values as a function of time were compared with the steady state anisotropy of a 0.5 μM sGFP2
solution without peptide (average value represented by the red line in Figure 2.15c, actual data
and error were omitted for clarity of the graph) and with the steady state anisotropy of monomeric
flGFP2 (average value represented by the blue line in Figure 2.15c). The purpose of these kinetic
anisotropy measurements was to observed if the incubation with M3 peptides induces changes in
the dimer:monomer ratio of sGFP2 by inducing the formation of more monomers. As shown in
Figure 2.15c, where 0.5 μM sGFP2 was incubated with 600 μM of M3 peptides, a decrease in
anisotropy was observed overtime and could be fitted by a single exponential decay with:

𝑦 = 𝐴 1
𝑒 (−𝑘 𝑜𝑏𝑠𝐴 𝑥 )
+ 𝑦 0
2.12
89

A plot of k obsA as a function of M3 peptide concentration indicated that k obsA saturates at
high peptide concentrations, suggesting that the decrease in anisotropy becomes rate-limited with
a rate constant of 0.01 min
-1
(Figure 2.15d). Interestingly, this rate constant is similar in value
with the rate limiting constant k obs2 observed during the complementation kinetics (Figure 2.15b).
Although, such rate-limiting decrease in anisotropy appears consistent with the expected effects
that the forward constant of dimer dissociation k f might have on the formation of new monomers
and the appearance of GFP fluorescence, k obsA could not be assigned to k f with confidence.
Indeed, a careful SEC analysis of the ratio between sGFP2 dimers and monomers after 13-14
hours incubation with large excess of complementary M3 peptides revealed that there was little-
to-no change in the respective amount of dimeric and monomeric sGFP2 after complementation
(not shown). Thus, while M3 peptides can bind sGFP2 monomers to form complemented and
matured full-length GFP2, they do not significantly impact the equilibrium between dimers and
monomers, at least within the time frame of our experiments. This suggests that k obsA/k obs2 is not
k f. Rather it appears to represent a rate-limiting constant, which we called k conf and that involves
an M3 peptide-dependent conformational changes of sGFP2 monomers that takes place before
maturation. Indeed, as shown in Figure 2.15c, the changes in anisotropy value over time never
fully reach that expected for a monomeric flGFP2. There is however a clear change in anisotropy,
which suggests a modification in the conformation of a sGFP2 species that can bind the M3
peptide, most likely monomers. It is possible that, under large excess of M3 peptides, partial
binding of multiple peptides to the open groove of a split-FP monomer could force the reaction
towards a much slower complementation pathway. Affected monomers with different number of
nonspecific M3 peptide interactions might undergo conformational change as a reverse reaction
step before maturation and appearance of fluorescence (Figure 2.16). This alternative
90

complementation pathway requires more measurements to fully understand how it affects the sFP
structure. By default, the slowest k obs3 rate was assigned to the forward constant of dimer
dissociation kf, since its value corresponds to an expected very slow rate of monomer formation,
as observed by SEC HPLC.
To sum up, we found that split-FPs exist in concentration-dependent monomer-dimer
equilibrium and only monomeric split-FP fragments are able to go through a rather slow and
irreversible binding to the M3 peptide fragment. This monomer-dimer equilibrium participates to
the observed differences in kinetic behaviors when complementation is performed within either
an excess of split-FP or an excess of M3 peptide fragments. While sGFPori has a faster k on
binding rate constant to the M3 peptide, it has a clearly slower k mat maturation rate constant
compared to other split-FPs.

91

Figure 2.15 Analysis of sFP chromophore formation kinetic based on the condition of M3 peptide titration. (a) Linear
function fit on kobs1 rate constants from each kinetic acquisition with different concentration of M3 peptide. The slope
defines the kon rate constant of binding step, and the y-intercept represents the kmat rate constant of maturation step. (b)
kobs2 and kobs3 rate constants from each kinetic acquisition as a function of different M3 peptide concentrations. kobs2 and
kobs3 rate constants are assigned as kconf (conformational change of monomeric sGFP2) and kf (forward reaction rate of
monomer-dimer equilibrium), respectively. (c) The fluorescence anisotropy kinetic measurements of sGFP2 only (red
line), sGFP2 with excess 600 μM M3 peptide, and flGFP2 only (blue line). The single exponential decay function was
used to fit the anisotropy kinetic values (green line). (d) Observed rate constants of each anisotropy kinetic
measurement, kobsA, from sGFP2 with different concentrations of M3 peptide.

92

Figure 2.16 Potential model of split-FP chromophore formation mechanism. In the condition of sFP titration (green
dashed rectangle), the binding step is followed by the maturation step. However, in the condition of M3 peptide
titration, after the binding step, the excess M3 peptide forces the reaction towards a yet not fully defined pathway that
requires a slow conformational change step on monomeric structures before proceeding to maturation. The
conformational change step is followed by the maturation step.

Live cells expression and imaging of split-FP variants
The development of novel split-FP variants provides a beneficial use for
complementation activated light microscopy (CALM) imaging of cellular processes [168].
CALM is a single molecule imaging technique that relies on the irreversible binding of split-FP
fragments and stochastic formation of the chromophore. As a first step to evaluate the potential of
some of our split-FPs for CALM imaging, we first tested their expression and complementation
in cells. U2OS cells were transiently transfected with cDNA encoding for different split-FPs
fused to the glycosylphosphatidylinositol anchoring domain (GPI) of the plasma membrane
protein CD14 (Figure 2.17). Different sFP-GPI fusions were imaged by ensemble confocal
microscopy upon complementation with large excess of the small M3 peptide fragment. As
shown in Figure 2.18, specific binding of the synthetic M3 peptide fragment to cells expressing
sGFPori-GPI, sGFP2-GPI, sGFP3-GPI, or sYFP3-GPI, resulted in fluorescence activation of
these different fusion proteins at the cell plasma membrane, indicating that complementation and
93

cell localization are not impeded and that these split-FPs can effectively be used for CALM
imaging.

Figure 2.17 The model for GPI anchored protein fusion with split-FP.

Figure 2.18 Confocal imaging of GPI anchored split-FP fusion transfected U2OS cells. The fluorescence images of the
cells (top panel) and bright field images (bottom panel) were shown. Scale bar: 15 μm.

The ability to complement efficiently the sFP-GPI fusions in cells indicates that there is a
substantial amount of monomeric sFP-GPI at the plasma membrane. Indeed, as defined in our in
vitro complementation assays, only monomeric sFPs can bind M3 peptides to give full-length and
fluorescent FPs. This observation, however, does not exclude the possibility that the sFP-GPI
94

fusions might still be in a dimer:monomer equilibrium in cells. This is an important factor to
consider in order to interpret adequately the diffusive behaviors and the cell membrane
distributions of the GPI-fusions. Indeed, dimeric fusions might have significantly different
biophysical properties than monomeric fusions at the cell surface. In order to assess the
oligomeric state of the sFP-GPI fusions at the cell plasma membrane, we therefore performed
cross-linking assays with the small amine-reactive bifunctional cross-linker
bis(sulfosuccinimidyl)suberate (BS3) [213]. BS3 is an irreversible and 1.2 nm long cross-linker
that non-specifically cross-links proteins by reacting with exposed amine groups. Because of its
small size, it can be used to cross-link proteins that are in very close proximity to each other and
it has effectively been employed to study the formation of dimers in cells [214]. Indeed, if a given
protein dimerizes in cells, dimers can be enriched after BS3 crosslinking and further visualized by
denaturing SDS page electrophoresis and immunoblotting. Using a U2OS cell line stably
expressing sGFP2-GPI, we incubated cells with 2 mM BS3 for 10 min, fixed them with 4%
paraformaldehyde (PFA) and homogenized them before denaturing SDS page electrophoresis and
immunoblotting with an anti-GFP antibody (Figure 2.19). The disulfide bond reducing agent N-
ethylmaleimide (NEM) was included in some of the reactions to verify that disulfide bond
formation between denatured sGFP2 (which contains three cysteines) did not induce post-
extraction multimerization. Under conditions where no BS3 or NEM is applied, sGFP2-GPI is
detected at about 32 kDa, a molecular weight similar to that expected for the fusion (theoretical
molecular weight is 32 kDa) (Figure 2.19). Notice that although the protein band is rather wide
because of gel retardation effects due to the GPI-anchor, there is no other higher molecular bands.
This indicates that sGFP2-GPI is primarily monomeric under denaturing electrophoresis
condition, as expected (Figure 2.19). The addition of NEM after extraction did not induce
changes in band position, indicating that there is little-to-no effect of disulfide bond formation on
95

the oligomeric state of sGFP2-GPI (Figure 2.19). However, when BS3 cross-linking was
performed, the monomeric band of sGFP2-GPI was accompanied with higher molecular weight
bands. Interestingly, these bands correspond to species with molecular weight in the range of 160-
200 kDa, much higher than expected for dimers of sGFP2-GPI. In fact, there was no GFP signal
detected at the expected ~64 kDa position for the sGFP2-GPI dimers (Figure 2.19). Again,
combining BS3 cross-linking with NEM treatment did not induce changes in band positions,
indicating that the high molecular weight bands of sGFP2-GPI likely stem from its random cross-
linking with other membrane proteins. Together these data demonstrate that sGFP2-GPI fusions
are only monomeric at the cell surface, with no dimer formed. This is a clear difference compared
to in vitro expressed sGFP2 and it implicates that the simple complementation model outlined in
Figure 2.19 is valid for sFP fusions in cells. Consequently, the complementation efficiency of sFP
fusions in cells and CALM imaging should only be dependent on the binding rate k on and on the
maturation rate k mat.

Figure 2.19 Immunoblot analysis of oligomerization of GPI anchored sGFP2 fusion on the cell plasma membrane.

We then tested and compared the properties of sGFPori and sGFP2 for single molecule
CALM imaging and tracking in live cells [168] (Figure 2.20). Cells expressing the sGFPori-GPI
96

or sGFP2-GPI were imaged during incubation with the small M3 peptide fragment to compare
brightness, photostability and complementation rates of both fusions in vivo.

Figure 2.20 The model of complementation activated light microscopy (CALM) for the complementation of sGFP-
transmembrane fusion protein and M3 peptide fragment on the cell plasma membrane [168].

As shown in Figure 2.21a for the sGFP2-GPI fusion, CALM imaging by total internal
reflection fluorescence (TIRF) excitation at 488 nm resulted in the appearance of individual
complemented and activated full-length GFP2-GPI fusions at 520 nm after addition of a low
concentration of M3 peptide to expressing cells. Individual GFP2-GPI fusion appearing at the
plasma membrane were localized by two-dimensional Gaussian fitting of their diffraction-limited
point-spread function, and their diffusion trajectories were reconstructed by linking the localized
position of each molecule from frame to frame (Figure 2.21a).
To compare the photophysical properties of individual and complemented sGFPori-GPI
and sGFP2-GPI in cells and under similar imaging conditions, cells co-expressing sGFPori-GPI
and the cyan nuclear marker LacI-CFP fusion were co-cultured with cells co-expressing sGFP2-
GPI and the red nuclear marker LacI-RFP. The cyan and red nuclear markers were used to
identify cells expressing sGFPori-GPI or sGFP2-GPI within a same field of view and single
molecule CALM tracking was performed following the addition of excess M3 peptides to prevent
variations in complementation kinetics due to possible different cellular expression levels
between both fusions. The difference in brightness between individual complemented sGFPori-
GPI and sGFP2-GPI fusions are shown in Figure 2.21b. As can be seen, a significant fraction of
97

individual sGFP2-GPI displays brighter fluorescent signals than sGFPori-GPI, consistent with the
fact that sGFP2 is 15% brighter than sGFPori because of its higher quantum yield, as determined
in bulk measurements (Table 2.1). In addition, we compared the length of trajectories at the cell
membrane for individual activated sGFPori-GPI and sGFP2-GPI fusions (Figure 2.21c). Again,
the trajectory length of sGFP2-GPI was systematically longer than that of sGFPori-GPI (Figure
2.21c). While this might suggest that sGFP2 is more photostable that sGFPori, such a hypothesis
is not compatible with our ensemble photobleaching measurements that indicate that both
proteins have similar irreversible photobleaching properties. However, the single molecule
tracking data reflect the fact that sGFPori is much more likely to be trapped in the
photoconvertible dark state compared to sGFP2. Indeed, entry in this dark state can result in
temporary loss of fluorescence signal (blinking) and in the inability to track a given molecule
between consecutive frames over long time periods.
We then compared the time of fluorescence appearance between sGFPori-GPI and
sGFP2-GPI in cells. To do so, single molecule events were recorded every 5 minutes over one-
minute long acquisitions spaced by 4 min intervals without excitation to allow fluorescence
complementation to proceed. Data were reported in percentage of newly detected trajectories at
different time points within a 15-45 min time window after addition of the M3 peptide, to allow
sufficient incubation and mixing of the peptide with cells. Reporting the percentage of newly
detected events allows a normalization of possible differences in expression level between cells.
As shown in Figure 2.21d-e, the slope of newly activated FPs over time is much shallower for
sGFP2-GPI than for sGFPori-GPI, indicating that a larger proportion of sGFP2-GPI has already
been complemented between 0-15 min and that only limited amounts of new molecules appear in
the 15-45 min window post incubation with M3 peptides. This is fully consistent with the fact
that sGFP2 has a 20 minutes faster maturation rate than sGFPori. Indeed, under similar imaging
98

and complementation conditions, only a difference in maturation kinetics would lead to the
observed difference in the proportion of newly activated molecule between sGFPori-GPI and
sGFP2-GPI in cells. Thus, compared to sGFPori, sGFP2 provide significant improvements for
single molecule CALM imaging in live cell because of its higher brightness, lower probability to
be trapped in the photoconvertible dark state, and overall faster rate of fluorescence appearance
after complementation.
99

Figure 2.21 CALM imaging and diffusion trajectories of single molecule GPI anchored split-GFP2 protein fusions. (a)
During the addition of M3 peptide, individual molecules in single frame upon complementation with M3 peptide
fragment (left), superresolved localizations of individual GPI anchor split-FP molecules (center), diffusion trajectories
of individual molecules. Scale bar: 10 μm. (b) Brightness comparison of individual sGFPori and sGFP2 at the plasma
membrane. (c) Trajectory length comparison of individual sGFPori and sGFP2 at the plasma membrane. (d-e) Time of
first appearance of complemented sGFPori and sGFP2 fragments upon introducing M3 peptide fragment.

100

Conclusion
Split fluorescent protein fragments are gaining increasing attention because they provide
many advantages for a variety of applications such as protein tagging, nanoassembly formation,
single molecule analysis, and neuronal imaging. Here, we have successfully developed better
folder and faster maturing sGFP variants of the original sGFPori as well as new sYFP and sCFP
spectral variants. A comparison between the full-length FPs and their split forms indicated that
they exhibit the same spectral features with identical maximum absorption and emission
wavelengths. The various site-directed mutations we have implemented in sGFPori led to
development of sGFP2, which has a 14% higher quantum yield than sGFPori. A similar
improvement of 17% was observed between the full-length flGFP2 and flGFPori. In addition to
the quantum yields, extinction coefficients have been evaluated for all FPs, leading to the
assessment that flGFP2 is the brightest of all the flGFP variants, followed by flGFP3, flGFP1 and
flGFPori. In the case of sGFPs, sGFP2 is the brightest followed by sGFPori, sGFP1, and sGFP3.
For the YFP variants, flYFP1 was the brightest followed by flYFP2 and flYFP3, while sYFP1
was brighter than sYFP3 and sYFP2. For the CFP variants, which were the least bright of all FPs,
flCFP2 and sCFP2 were brighter than flCFP1 and sCFP1 respectively. The typical fluorescence
lifetime of the GFP variants was single exponential and in the range of 2-3 ns, an observation
consistent with a stabilized B-state and benzoidal form of the chromophore. The lifetime decays
of the CFP variants were biexponential with participation from the chromophore A- and B-states,
as expected. Two of the YFP variants also displayed two fluorescence lifetimes originating from
a participation of the I- and A-states of the chromophore, respectively. We found, however, that
the YFP1 variant displays a peculiar and single lifetime decay, due to a stabilization of the
chromophore B-state that is induced by a combination of T65G and T203Y mutations.
Interestingly we also found that split-FPs systematically have ~200 ps shorter lifetime than the
101

full-length FPs. A potential explanation for this result is that the chromophore in split-FPs is
likely more exposed to the surrounding medium upon complementation with the M3 peptide
fragment than the same chromophore in full-length FPs.
With respect to photostability, the different mutations implemented in sGFPori led to
clear improvements in irreversible photobleaching kinetics, notably for the sGFP1 and the sGFP3
variants and in a reduced probability to be trapped in a photoinduced reversible dark state. The
same mutations also significantly improved the folding and the chromophore maturation rate of
the original GFPori. In particular, mutations implemented in sGFP2 led to a 20 min increase in
the maturation rate constant of sGFPori.
Through carefully designed complementation and maturation kinetic experiments, we
also demonstrated that the large split-FP fragment exists in a concentration-dependent monomer-
dimer equilibrium in vitro and that only monomeric split-FPs are capable of binding M3 peptides
via a conformational selection mechanism. The binding rate k on is rather slow but irreversible,
and it is followed by a rapid maturation rate of the chromophore. In this process, we showed that
sGFPori binds M3 peptide slightly faster than sGFP2, but that its chromophore maturation rate is
nearly 50 % slower than for sGFP2. Surprisingly, this simple complementation kinetics is
impacted by the presence of large excess of M3 peptide fragment which appears to drive a slower
alternative complementation pathway involving a conformational change of the affected split-FP
monomers. A partial binding of multiple M3 peptides at the missing 11
th
beta-sheet site in split-
FP, might explain why such alternative complementation pathway is triggered.
We also demonstrated that, GPI anchored fusions to sGFPori, sGFP2, sGFP3, and sYFP3
are properly expressed in live cells and can be efficiently complemented with M3 peptides for
single molecule and ensemble confocal fluorescence microscopy imaging. Unlike split-FPs
expressed in vitro, split-FP-GPI fusions exist only as monomers at the cell plasma membrane.
102

When we compared the biophysical properties of sGFPori and sGFP2 fusion proteins in live cells
by single molecule imaging and tracking, we found that sGFP2-GPI fusions mature faster, are
brighter and allow longer single molecule tracking and trajectory reconstruction than sGFPori-
GPI fusions. These improvements are due to the faster maturation rate, the higher quantum yield,
and the lower propensity of sGFP2 to be trapped in a light-induced photoconvertible dark state
compared to sGFPori. In conclusion, sGFP2 provides a good balance between fast maturation
rate, efficient binding rate, high brightness and good photostability that makes it an excellent
candidate for future ensemble and single molecule fluorescence imaging by CALM, and for
applications involving the controlled assembly of nanomaterials using split-FP fragment
complementation.

Material and Methods
Expression and purification of FPs
A Quickchange lightning site-directed mutagenesis kit (Agilent Technologies) and
appropriate template primer were used to make site-directed mutations and design the spectral
variants of sGFPori and flGFPori. DH5α competent cells were used to transform the plasmids
with kanamycin resistant cassette. A BL21(DE3) E. coli strain was used to express the different
protein variants. All FPs have been expressed with an N-terminal tetracysteine tag and with an N-
terminal 6xHis-tag for purification. A detailed protocol for expression and purification is
available in CHAPTER 3, section 3.8.1 - Expression and purification of sGFP.

Spectral acquisitions and photobleaching kinetics of FPs
For complementation, purified split-FPs (20 μM) and M3 peptide (200 μM, sequence
available in CHAPTER 3, section 3.8.1 - Expression and purification of sGFP) were incubated
103

for 12 hours in TNG buffer (100 mM Tris-HCl, 150 mM NaCl, 10% glycerol, pH 8.0) at RT.
Full-length FP variants were diluted at 5 μM in TNG buffer. A Varian Cary® 50 UV/Vis was
used to acquire absorption spectra of all FPs. A Horiba Nanolog spectrofluorometer was used to
acquire emission spectra of all the FPs and for photobleaching kinetics. For photobleaching
kinetics, complemented sGFP variants in TNG buffer were constantly excited at 488 ± 1 nm and
fluorescence emission was collected at 530 ± 1 nm for 40 minutes at RT. Photobleaching kinetic
data were analyzed using the model in Figure 2.9a. The results were fitted by the solution to the
differential equations described in Equation 2.6 using the software Matlab.

Fluorescence lifetime measurements
We have used fluorescence lifetime imaging microscopy (FLIM) technique in frequency
domain to determine the lifetime of split and full-length fluorescence proteins in TNG buffer (100
mM Tris-HCl, 150 mM NaCl, 10% glycerol, pH 8.0). A Zeiss LSM 780 inverted microscope
equipped with 40X water immersion objective (NA 1.1) was used for FLIM measurements.
Samples were excited by a two-photon laser at 870 nm (5-10 mW) (Chameleon, Coherent) with
150-fs pulse bandwidth at 80 MHz repetition rate. 537/26 nm bandpass emission filter was used
for GFP/YFP, and 483/32 nm bandpass emission filter was used for CFP. Emitted light intensity
was measured by two hybrid photodetectors (Hamamatsu) where the acquisition time was
adjusted to achieve an average of 100 counts/pixel.
The lifetime values were determined by one or two-component exponential fit of the
fluorescence decay curves using an ISS VistaVision software (version 4.1).

104

Refolding and maturation kinetics of full-length FPs
To measure the refolding kinetics, 50 μM of full-length FP variants in TNG were
denatured by the addition of 8M urea and boiled at 95 ˚C for 10 minutes. 1mM DTT and 0.2 mM
EDT were also used to reduce potential disulfide bonds. Denatured samples prepared in triplicates
were diluted by 100-fold in TNG on a 96-well microplate to start the refolding. A Biotek Synergy
H4 microplate reader equipped with a xenon lamp, excitation filters at 452/17 nm, 485/20 nm or
500/13.5 nm and emission filters at 480/9 nm, 528/20 nm or 526/9 nm were used to acquire the
refolding kinetics in 25 seconds increments for 2 hours at 25 ˚C. Microplate wells containing
TNG buffer or the appropriate non-denatured full-length FP variants were used for buffer
background corrections and for photobleaching corrections. As mentioned in the main text,
refolding kinetics were analyzed by a triexponential function (Equation 2.7) using Origin 2016 as
a software.
For maturation kinetic measurements, 5 μM of full-length FP variants in TNG were
denatured by the addition of 8M urea and boiled at 95 ˚C for 10 minutes in the presence of 5 mM
dithionite to reduce the chromophore. 1mM DTT and 0.2 mM EDT were also used to reduce
potential disulfide bonds. The maturation kinetic was triggered by a 100-fold dilution of triplicate
FP samples in TNG buffer and data were acquired on the Biotek Synergy H4 microplate reader as
described above. Full-length FP samples were prepared without urea, dithionite or boiling for
long-term photobleaching corrections. Maturation kinetics were fitted by a single exponential
function (Equation 2.8) using Origin 2016 as a software.

Self-assembly kinetic of split-FPs
For all FPs, a Biotek SynergyH4 microplate reader equipped with appropriate excitation
and emission filters was used to acquire fluorescent signals at 25 ˚C, every 3 minutes over a 14
105

hours period. M3 titrations were prepared in triplicate using 12 different concentration points;
0.05 μM, 0.12 μM, 0.28 μM, 0.65 μM, 1.52 μM, 3.57 μM, 8.4 μM, 19.72 μM, 46.31 μM, 108.76
μM, 255.47 μM, and 600 μM of M3 peptides in 1 mM DTT, 5 mM EDTA pH 8.0, and 0.05 %
CHAPS. 0.5 μM sFP in TN buffer (100 mM Tris-HCl, 150 mM NaCl, pH 8.0) was equilibrated
for at least 4 hours at RT. M3 peptides were added to the 96 well plate right before the start of the
acquisitions. Samples without sFP and without M3 peptides were used for buffer background
correction. A 0.1 μM of corresponding full-length FP was used for long-term photobleaching
corrections. Kinetic curves were fitted by a tri-exponential function (Equation 2.11) using Origin
2016 as a software.
sFP titrations were preformed using 12 different sFP concentration points; 0.1 μM, 0.17
μM, 0.28 μM, 0.47 μM, 0.8 μM, 1.34 μM, 2.24 μM, 3.77 μM, 6.33 μM, 10.64 μM, 17.86 μM and
30 μM of total (dimer and monomer) sFP in TN buffer with 1 mM DTT, 5 mM EDTA pH 8.0,
0.05 % CHAPS. sFP samples were prepared in a serial dilution from a main stock (37 μM), which
was equilibrated for at least 4 hours at RT before the kinetic measurement. 0.1 μM M3 peptide
was added right before the start of the acquisitions. Three replicate wells were prepared for each
concentration point. Samples without sFP and without M3 peptides were used for buffer
background correction. 0.1 μM solution of corresponding full-length FP was used for long-term
photobleaching corrections. Kinetic curves were fitted by a biexponential function (Equation
2.10) using Origin 2016 as a software.

Anisotropy measurements
For all the anisotropy measurements, a Biotek SynergyH4 microplate reader equipped
with appropriate vertical and horizontal polarization filters, a 540/25 excitation filter and a 620/40
emission filter was used. Labeling of sGFPori or sGFP2 was done with a 1:2 molar ratio of sFP to
106

ReAsH dye (40 μM:80 μM) in the presence of 5 mM TCEP and 2 mM BME for 1.5 hour at RT.
The excess of ReAsH dye was removed using a Sephadex gel filtration G-10 spin column
(Harvard Apparatus).
For dimer:monomer Keq studies, endpoint fluorescent anisotropy measurements were
acquired from 6 different concentrations of ReAsH-sFP (20, 10, 5, 2.5, 1.25 and 0.5 μM) in
triplicate, after 40 hours equilibrium at RT. Keq values were determined by fitting the anisotropy
as a function of sFP concentration with the equation of Figure 2.13, using Origin 2016 as a
software.
For M3 peptide induced changes in anisotropy kinetics, 0.5 μM ReAsH-sFP was
equilibrated for 5 hours at RT. M3 peptides were prepared at 12 different concentration points in
triplicate: 0.05 μM, 0.12 μM, 0.28 μM, 0.65 μM, 1.52 μM, 3.57 μM, 8.4 μM, 19.72 μM, 46.31
μM, 108.76 μM, 255.47 μM, and 600 μM. M3 peptides were added to ReAsH-sFP immediate
before the start of the acquisition. The anisotropy was measured every 3 min over a 14 hours
period. The anisotropy decay over time was fitted by a single exponential decay function
(Equation 2.12) to extract K obsA at each M3 peptide concentrations using Origin 2016 as a
software.

Cell lines, cell labeling, and confocal imaging
U2OS cells were maintained in DMEM (Lonza) supplemented with 10% fetal bovine
serum (FBS, Gibco) in a humidified incubator at 37ºC, supplied with 5% CO2. Cells were
transiently transfected with cDNA encoding for the different sFP-GPI fusion using Roche,
XtremeGENE HP transfection reagent.
Ensemble confocal fluorescence images were acquired on a Nikon C2 inverted confocal
microscope using a 488 nm excitation laser and a 525/50 nm band-pass emission filter (Semrock)
107

for complemented sGFP variants, or a 515 nm excitation laser and a 542/27 nm band-pass
emission filter (Semrock) for complemented sYFP variants. Cells transiently transfected with the
different sFP-GPI fusions were imaged after the addition of 37 μM M3 peptide in the DMEM +
10% FBS culture media for 12 hours. Cells were briefly rinsed with HBSS buffer (Corning) and
imaged in the same buffer at 37°C.
For comparative single molecule imaging and tracking, U2OS cells were transiently co-
transfected for 48 hours with sGFPori-GPI and LacI-CFP or sGFP2-GPI and LacI-RFP in
separate plates. Cells were then gently detached using Cellstripper (Corning), mixed, and plated
on the same microscope coverslip for 24 hours at at 37°C prior to imaging. For imaging, cells
were rinsed 3x with 37ºC HBSS (Corning), and imaged by TIRF microscopy in HBSS buffer
after the addition of 45 μM M3 peptide.

Single molecule imaging and tracking
Imaging was performed on a Nikon Eclipse Ti-E microscope equipped with a 100x 1.49
NA objective (Nikon), a iXon EMCCD camera (Andor), a laser line at 488 nm (Agilent), a
multiband pass ZET 405/488/561/647x excitation filter (Chroma), a quad-band
ZT405/488/561/647 dichroic mirror, and a 525/50 nm bandpass emission filter (Semrock). TIRF
excitation was used to image the bottom of the plasma membrane for 7 minutes continuously at
an acquisition rate of 40 ms per frame after addition of 45 μM M3-biotin peptide, then for 1
minute at the 10 minute mark, and 1 minute at each additional 5 minute mark for up to 45
minutes.
Particle analysis was performed using SlimFast [215], a single-particle detection and
tracking software written in MATLAB that uses multiple-target tracing algorithms, provided by
Christian Ritcher and Jacob Piehler. Localizations were performed on individual molecules by 2D
108

Gaussian curve fitting of the point-spread function of each sGFP in each frame. Precision was
estimated using the algorithm of Thompson et al [216]. Diffusion trajectories were built by
linking individual localized events from frame to frame, taking into account blinking statistics
and local particle densities. Trajectories with at least 3 steps were kept for analysis.

Immunoblot analysis
U2OS cells stably expressing sGFP2-GPI fusions were used to assess the oligomeric state
of split-FP fusion in cells. Briefly, cells were rinse with PBS and incubated with 2 mM
bis(sulfosuccinimidyl)suberate (BS3) for 30 minutes at 4 ˚C, before quenching the cross-linking
reaction with 20 mM Tris pH 8.0 for 15 minutes. For NEM treatments, 20 mM N-ethylmaleimide
(NEM), a disulfide bond reducing agent, was applied on cells for 10 min at room temperature
after the Tris quenching steps. Cell were then immediately fixed with 4% paraformaldehyde
(PFA) and scraped from the petri dish before being lysed in a cell homogenizer. Cell extracts
were run on a denaturing SDS-page electrophoresis gel, before transfer on PVDF immunoblotting
membrane (Biorad) and blocking with 5% dry milk in TBST buffer (20 mM Tris-HCl, 150 mM
NaCl, pH 7.4 with 0.05 % tween-20). Detection of sGFP2-GPI was performed with a monoclonal
mouse anti-GFP (JL-8) primary antibody (Clontech 632381, 1:2000 dilution) and an HRP-
conjugated polyclonal goat anti-mouse secondary antibody (Invitrogen 31430, 1:3000 dilution)).
Photoluminescent signal were detected on a Biorad Chemidoc using an SuperSignal West Pico
chemiluminescent substrate (ThermoFisher).
A monoclonal mouse anti-GFP (JL-8) primary antibody (Clontech 632381, 1:2000
dilution) and a polyclonal goat anti-mouse secondary antibody (Invitrogen 31430, 1:3000
dilution)) were used for immunoblot analysis.

109

CHAPTER 3. TARGETED CELL IMAGING BY IN-SITU
ASSEMBLY AND ACTIVATION OF HOT SPOT SERS
NANOPROBES USING SPLIT-FLUORESCENT PROTEIN
SCAFFOLDS

Introduction
Noble metal gold (Au) and silver (Ag) colloidal nanoparticle (NPs) are particularly well
suited to design optical probes for advanced biodetection and bioimaging applications because
their nanoscale photophysical properties often surpass those of the best chromophores [217-219].
Their large optical cross-section, easy bio-functionalization and shape-tunable photo-response
across the visible and near-infrared spectra have opened new imaging capabilities by surface
plasmon resonance [51], photoacoustic detections [220] and surface enhanced Raman scattering
(SERS) [221]. When employed for SERS, plasmonic metal NPs provide highly sensitive optical
detections of the vibrational signatures of Raman reporters positioned at or near their surface
[222]. The strong near-field electromagnetic amplifications generated by optical excitation of
metal NPs can indeed overcome the intrinsically low Raman cross-section of absorbed molecules
and result in Raman scattering enhancement factors of 10
2
-10
12
folds [105, 223] depending on the
shape and the composition of NPs and on the number and the position of Raman reporters at their
surface.
For targeted cell imaging by Raman scattering, SERS nanotags consisting of a spherical
or anisotropic metal NP core pre-activated with thousands of surface Raman reporters are often
used [224-227]. Such high-density coatings of the reporters and additional encapsulation in
110

protective shells are required to compensate for the modest SERS enhancements of the NP core
(10
2
-10
5
folds) and to generate sufficient Raman signals for single cell [228] and in vivo
multispectral imaging [60, 93]. An alternative approach to engineer SERS nanoprobes having
greater detection sensitivity than SERS nanotags is the directed self-assembly of metal NPs into
dimers or higher order nanoclusters and the positioning of Raman reporters within interfacial
nanogaps between clustered NPs [229]. Interparticle plasmon-plasmon couplings between
clustered NPs produce massive near-field amplifications at nanogaps and the 10
8
-10
12
fold SERS
enhancements within these plasmonic hot spots enable single molecule SERS detections [4, 108,
230-232]. Such high enhancements are, however, strongly dependent on the stability of the
Raman reporters within hot spots and on the size of the interparticle gap [229, 233], which are
both rarely well controlled [234]. For nanogaps larger than 1-2 nm, near-field amplifications
decay rapidly due to reduced plasmon coupling between NPs [235] and for smaller nanogaps
electron tunneling and field dissipation lower SERS enhancements [236]. Forming nanosized
plasmonic hot spots reproducibly and precisely positioning Raman reporters at these sites is very
challenging and the development of hot spot SERS nanocluster probes for bioimaging remains
limited despite their significant advantages for ultra-sensitive detections [237].
A promising approach to generate SERS probes with multiple well-defined nanogaps and
pre-programmed hot spots is to employ Raman reporters that also act as molecular glue for the
controlled bottom-up assembly of clusters, for instance using host-guest interactions between
complementary molecules appended to the surface of different NPs [238]. This strategy has been
used to assemble NP SERS beacons, where nanoclustering driven by complementary nucleic acid
scaffolds enhances the Raman scattering of chromophores pre-encoded at the surface of NPs or
within the scaffold itself [239-242]. These approaches, however, suffer from multiple drawbacks,
including (i) background SERS or fluorescence signals from the reporters, which degrade high
111

signal-to-noise SERS detections, (ii) limited control of the nanogap size due to the lack of
structural rigidity of nucleic acid scaffolds and (iii) the difficulties to carry out such assemblies in
cells. Indeed, while nucleic acids remain the building blocks of choice for the in vitro bottom-up
assembly of colloidal NPs into photonic nanomaterials [80, 243], these scaffolds are degraded by
nucleases and are very susceptible to pH and ion concentrations in biological buffers. Specificity
and binding affinity are also anti-correlated in the one-dimensional zipping mechanism that
underlies selective interactions between nucleic acids, which further limit their use for the remote
assembly of SERS nanocluster probes in cells and in vivo.
These limitations, however, can be overcome by employing other bio-inspired scaffolds
such as self-complementary proteins and peptide fragments, whose secondary and tertiary
structures provide unmatched binding specificity and affinity [244]. Among those, fluorescent
proteins (FPs) offer many advantages for the controlled supramolecular assembly of metal NPs
into SERS nanocluster probes. They are preprogramed to rapidly self-assemble from highly
evolved protein domain building blocks and their folding mechanisms into compact and sturdy
nanoscale entities is well understood [111]. They also encode fully biocompatible peptide-based
chromophores that self-activate upon folding and whose peculiar Raman fingerprints can easily
be differentiated from the vibrational signatures of other biomolecules [112, 113] with single
molecule SERS sensitivity [100].
In this chapter, we demonstrate that split fragments of the green fluorescent protein
(GFP) and its yellow (YFP) and cyan (CFP) spectral variants can be used as molecular glue and
activatable Raman reporters to drive the assembly of metal NPs into activatable hot spot SERS
nanocluster probes having plasmonic nanogaps homogenously seeded by the precise positioning
of fully folded FPs. Split GFP fragments are complementary protein domains from a super-folder
GFP split into two highly asymmetric pieces [175]: a large GFP 1–10 domain (sGFP, amino acids
112

1–214) and a smaller M3 peptide domain corresponding the 11th β-sheet of the super-folder GFP
β-barrel structure (M3, amino acids 215–230). Both fragments, including synthetic versions of the
M3 peptide, spontaneously and irreversibly self-assemble in solution to form a fully folded GFP
within which the peptide-chromophore can mature [168, 245]. This bimolecular complementation
system has been used to target nanomaterials in cells [168, 246] and for the supramolecular
scaffolding of protein nanostructures [167]. Here, we show that, when appended to the surface of
different metal colloids, these FP fragments guide the self-assembly of NPs into active plasmonic
clusters and trigger the autocatalytic maturation of the chromophore within FPs fully
reconstituted at interfacial nanogaps. Upon dual-NP targeting to plasma membrane biomarkers in
live cells, these NPs rapidly form discrete nanoclusters and the activation of the FP chromophore
Raman signatures within plasmonic hot spots allows highly specific and site-directed SERS
imaging of cells. This in-situ and FP-assisted clustering of metal NPs also yields strong
photoacoustic signal amplifications, allowing the nanoclusters to serve as contrast agents for
multimodal SERS and photoacoustic imaging of individual cells.

Guided clustering of metal NPs with controlled nanogaps using FP fragments
To exploit split FP fragments as surface molecular glue for the assembly of metal NPs
(Figure 3.1a) we first functionalized 40 nm AuNPs with either the large sGFP fragment or its
complementary M3 peptide fragment.
113

Figure 3.1 Characterization of AuNPs functionalized with split-fluorescent protein fragments. (a) Schematic of surface
modification on AuNPs by sGFP and M3 peptide fragments and formation of SERS active hot spot through self-
assembly and GFP complementation. (b) Comparison of the absorption spectra of bare AuNPs (solid black), AuNPs
coated with M3 peptides (M3-AuNPs, solid red), and AuNPs coated with sGFP (sGFP-AuNPs, dash blue). Inset: Size
exclusion chromatography of M3-AuNPs. (c) Immuno-dot blot characterization of the presence of full-length GFP or
sGFP at the surface of AuNPs. (d) TEM images of monodispersed AuNPs coated with M3 peptides or sGFP. Scale bar:
200 nm. (e) DLS size distributions of bare AuNPs, M3-AuNPs and sGFP-AuNPs, coated with different size PEGs.

sGFP-AuNPs were obtained by ligand exchange on citrate-stabilized NPs using a recombinant
sGFP modified with a N-terminal tetracysteine motif [247] and a flexible linker domain that
114

allow its oriented binding and its conformational flexibility at the metal surface (Figure 3.2,
Figure 3.3, Figure 3.4).

Figure 3.2 Engineering sGFP for oriented binding on AuNPs. sGFP was expressed as a recombinant protein with: (i) a
6xHis-tag for purification of the protein, (ii) a GSS linker sequence, (iii) a thrombin cleavage site (LGLVPRC) to cut
out the 6xHistag with thrombin after purification, (iv) a tetracysteine motif for oriented binding of sGFP at the surface
of AuNPs, (v) a flexible GGSGG linker domain to limit conformation stiffness of the protein on AuNPs and (vi) the
sGFP fragment coding sequence.

Figure 3.3 SDS-PAGE gel electrophoresis characterization of the expression and the purification of the recombinant
sGFP fragment. Lane 1: Molecular weight ladder. Lane 2: Commercial sGFP (~25KDa). Lane 3: Unpurified cell lysate.
Lane 4: 6xHistag purified sGFP (~26KDa). Lane 5: Thrombin cleaved sGFP. A few higher molecular dimers of sGFP
are observed (9%).

115

Figure 3.4 Gel electrophoresis and ReAsh labeling to assess the presence and the activity of the tetracysteine motif at
the N-terminus of the sGFP fragment. Lane 1: sGFP. Lane 2: ReAsh labeled sGFP. Lane 3: sGFP + M3 peptide
fragment complementation. Lane 4: ReAsh labeled sGFP + M3 peptide fragment complementation. Notice that
intramolecular FRET between complemented GFP and ReAsh bound to the tetracysteine motif is observed in lane 4.

A split-YFP fragment (sYFP) and a split-CFP fragment (sCFP) were also employed as
substitutes of sGFP for coatings, and the full-length super-folder GFP (flGFP) was used as a
control. Complementary M3-AuNPs were functionalized with a synthetic M3 peptide fragment
modified with a terminal cysteine for high affinity binding to the metal surface [248]. For both
sets of AuNPs, thiolated PEG-biotin (molecular weights of 5000, 2000 or 600 Da) was added
during surface functionalization to increase the stability of the NPs. These coatings resulted in
small spectral shifts of the localized surface plasmon resonance peak for both M3-AuNPs (λ max:
528 nm) and sGFP-AuNPs (λ max: 528 nm) compared to bare, citrate-stabilized AuNPs (λ max: 524
nm); a manifestation of local changes in refractive index typically observed after surface grafting
of biomolecules [249] (Figure 3.1b). The stable surface anchoring of sGFP was confirmed by
immuno-blotting assays with anti-GFP antibodies on extensively purified sGFP-AuNPs and clear
immuno-reactivity was observed compared to citrate-stabilized AuNPs (Figure 3.1c).
Interestingly, the missing C-terminal 11th β-sheet in sGFP resulted in weaker antibody reaction
116

on sGFP-AuNPs than on AuNPs coated with flGFP (Figure 3.1c), suggesting that surface-bound
sGFPs are oriented with their C-terminus exposed towards the buffer and accessible for binding
complementary M3 peptides. The colloidal solutions of sGFP-AuNPs and M3-AuNPs were very
stable with no signs of aggregation during size exclusion liquid chromatography (inset of Figure
3.1b) and gel electrophoresis (Figure 3.5) and the NPs were monodispersed in transmission
electron microscopy (TEM) images (Figure 3.1d and Figure 3.6). Consistent with the expected
influence that different FP fragments and different PEGs might have on the hydrodynamic
diameter of functionalized NPs, the size of citrate-stabilized AuNPs (51±4 nm) increased to 51-62
nm for M3-AuNPs and 60-65 nm for sGFP-AuNPs, depending on the length of PEGs employed
during coating (Figure 3.1e and Figure 3.7). Thus, various split FP fragments could readily be
grafted and oriented at the surface of AuNPs with minimal impact on their final size, their
photophysical properties and their colloidal stability.

Figure 3.5 Agarose gel electrophoresis of bare, citrate-stabilized AuNPs (lane 1), sGFP-AuNPs (lane 2), and full-length
super-folder GFP-coated AuNPs (lane 3). The direction of the electric field applied is indicated by the arrow. Notice
that bare AuNPs are unstable under these electrophoresis conditions and aggregate shortly after entering the gel.

117

Figure 3.6 TEM images of AuNPs functionalized with split-fluorescent protein variants sYFP and sCFP. Scale bars:
200 nm.

Figure 3.7 Dynamic light scattering characterization of AuNPs coated with sGFP or M3 peptide fragments and with
different sizes of biotin-PEG moieties (5000 Da, 2000 Da, and 600 Da).

To determine if the spontaneous self-assembly of M3 and sGFP fragments would
effectively induce the clustering of AuNPs, we simply co-incubated different sizes of M3-AuNPs
and sGFP-AuNPs in buffer solutions. After 12 hours, clusters were detected as smeared AuNP
bands positioned at varying location in electrophoresis gels depending on the size of NP used in
118

the reaction (40, 20 or 10 nm AuNPs, Figure 3.8a, Figure 3.9). Typically, smears were
accompanied by a lost, sometime only partial, of one or both M3-AuNP and sGFP-AuNP bands.
When these smears were electro-eluted from gels, AuNP clusters with various length and shapes
were observed by TEM, including AuNP homo- and heterodimers, AuNP chains, and more
complex two-dimensional and three-dimensional AuNP assemblies (Figure 3.8a).

Figure 3.8 Formation of AuNP clusters by the assembly of FP fragments. (a) Agarose gel electrophoresis of sGFP-
AuNPs, M3-AuNPs and AuNP clusters and typical TEM images of assembled clusters. White rectangles indicate the
position of cluster bands in gels. Scale bars: 20 nm. (b) TEM images of clusters formed by the assembly of M3-AuNPs
with sGFP-AuNPs, sYFP-AuNPs, or sCFP- AuNPs. Scale bars: 200 nm. (c) Normalized AuNP cluster size distribution
and fit by a power law distribution model. (d) Size distribution of nanogaps formed by the assembly of FP fragments
between AuNPs in clusters. The distribution is Gaussian and centered at 2.1±0.5 nm. Inset: expected orientation of
complemented GFP at nanogaps. (e) TEM images of clusters formed by the assembly of sGFP-gold nanorods with
different sizes of M3-AuNPs. Scale bars: 20nm.

119

Figure 3.9 Agarose gel electrophoresis of 40 nm citrate-coated AuNPs functionalized with sGFP and co-incubated with
10 nm oleic acid-coated AuNPs functionalized with M3 peptides. TEM images of some of the nanoclusters observed
after gel extraction of the gel band corresponding to the nanoclusters (dash lines). Scale bar: 20 nm.

The clustering process was solely dictated by the bimolecular complementation between
the surface-attached FP fragments as confirmed by direct competition with an excess of free and
non-cysteinilated M3 peptides (Figure 3.10). The complementation-driven assembly of AuNP
clusters was observed using sGFP-AuNPs but also with the sYFP-AuNPs and sCFP-AuNPs
spectral variants (Figure 3.8b and Figure 3.11), an indication that the clustering process is broadly
applicable to different sizes of AuNPs and to different types of split FP fragments.

Figure 3.10 Gel electrophoresis competition assay of the formation of AuNPs nanoclusters using an excess of free and
non-thiolated M3 peptide fragment. Upon 12 hours co-incubation of 10 nm sGFP-AuNPs (lane 1) with 40 nm M3-
AuNPs (lane 2), a typical smeared band of AuNP nanoclusters is observed together with the disappearance of the M3-
AuNP band (lane 3). If a large excess (100 µM) of free, non-cysteinilated M3 peptide fragment is added at the
120

beginning of the co-incubation, little to no smearing is observed and both sGFP-AuNP and M3-AuNP bands remain
intact (red arrow heads). This indicates that the formation of AuNP clusters is solely driven by the self-assembly of
split fluorescent protein fragments appended at the surface of the nanoparticles.

Figure 3.11 TEM images of compact nanoclusters formed after co-incubation of sGFP-AuNPs, sYFP-AuNPs, or sCFP-
AuNPs with M3-AuNPs. Scale bar: 200 nm.

Using 40 nm AuNPs, we also studied the kinetic of cluster formation. While AuNP
clusters were formed within an hour of co-incubation (Figure 3.12), the clustering kinetic was
slow and at least 12 hours co-incubation of M3-AuNPs with sGFP-AuNPs at room temperature
was required to induce a clustering of more than 50% of all the NPs (Figure 3.13). Consistent
with the competition gel assays, this slow kinetic suggested that binding and steric fit between
complementary FP fragments, rather than diffusion-driven collisions of the NPs, is the primary
mechanism of cluster formation. To characterize further this assembly process, we performed
statistical analyses on the size heterogeneity of the AuNP clusters, which critically depends on
whether the kinetic regime of clustering is diffusion- or reaction-limited [250, 251]. After 12
hours, 53% of all AuNPs formed clusters with at least two NPs. The assembly process was not
random because the cluster size distribution could not be adequately described by a Poisson
distribution model (Figure 3.13). However, it followed a power-law probability distribution with
an exponent of 1.7±0.3 (Figure 3.8c), as typically observed for irreversible and reaction-limited
121

clustering of colloids in solution [250, 251]. Thus, the slow formation of predominantly small and
compact AuNP clusters observed by TEM (Figure 3.8b and Figure 3.11) indicates that the
clustering kinetic of M3-AuNPs with sGFP-AuNPs is primarily reaction limited, as expected for
an assembly driven by the irreversible bimolecular complementation between M3 and sGFP
fragments. This relatively slow kinetic reflects the importance of steric fit between the
complementary FP fragments, which have reduced conformational flexibility at the surface of
AuNPs and whose molecular interactions in solution are likely impacted by the three-dimensional
degrees of freedom of the diffusing NPs. We note that, while ~12 hours incubation is required to
obtain 53% of clustered AuNPs, the bimolecular reaction of sGFP2 with M3 peptide fragments
freely diffusing in solution reaches saturation within ~4 hours, as shown in CHAPTER 2. The
slow clustering rate of AuNPs further supports the idea that the complementation follows a
reaction-limited process due to the steric constraints that surface appended sFP fragments
encounter once bound to AuNPs; constraints that are not present when the same fragments diffuse
freely in solution. These increased steric constrains are also accompanied by a slower diffusion of
the AuNPs compared to the free fragments due to the nanoparticle much larger size. Although
this additional effect of slower diffusion might also participate to reducing the complementation
kinetics between the sFP fragments, the size difference between AuNPs and sFP fragments is not
such that it would affect the complementation kinetics as dramatically as observed here. In other
words, the reduced conformational flexibility of sFP fragment at the surface of AuNPs is likely
the primary reason for the slow clustering kinetics of AuNPs.

122

Figure 3.12 Kinetic analysis of the formation of AuNP clusters in solution by dynamic light scattering (DLS). M3-
AuNPs-PEG600 (40 nm) were co-incubated with sGFP-AuNP-PEG600 (40 nm) in solution. The formation of AuNP
clusters was assessed by measuring the changes in the size distribution of the AuNPs using DLS measurements at
different time intervals (1, 2, 4, 6, 24 and 48 hours) and comparing with their initial size distribution at time t=0 min.

123

Figure 3.13 Statistical size distribution of unpurified AuNP clusters observed by TEM and formed by co-incubating
sGFP-AuNPs (40 nm) with M3-AuNPs (40 nm) for 12 hours (red bars). The cluster size distributions were determined
over five independent assembly experiments. 53% of all the AuNPs assemble into clusters containing at least two
AuNPs. The distribution is compared to the expected Poisson distribution if the formation of the clusters was driven by
random interactions between AuNPs (black line, λ = 0.53).

A salient characteristic of this FP-based assembly strategy is that the interfacial gap size
between clustered AuNPs should be uniform and controlled by the orientation of the
complemented FP once split FP fragments are reassembled. To assess if this was indeed the case,
we measured the dimension of more than 300 nanogaps between clustered AuNPs from TEM
images. The resulting gap size histogram was well described by a Gaussian distribution centered
at 2.1±0.5 nm (Figure 3.8d and Figure 3.14), a size that corresponds to the short axis of the GFP 2
nm x 4 nm cylindrical structure [252] and is consistent with a transverse orientation of
complemented GFPs at nanogaps (Figure 3.8d). This observation, together with SERS
measurements discussed below, indicates that the interparticle spacing within clusters is governed
by the folding of the two split FP fragments into a re-assembled GFP at the interface between
AuNPs. The formation of stable colloidal clusters with nanogap dimensions controlled by the
self-assembly of FP fragments was also efficient for a variety of other nanomaterials, including
gold nanorods (Figure 3.8e) and AgNPs (Figure 3.15). Thus, split FP fragments appended to the
surface of metal NPs effectively act as molecular glue to guide the self-assembly of clusters
having well-defined and small nanogaps for a variety of plasmonic nanomaterials.

124

Figure 3.14 Nanogap size measurement between two 40 nm AuNPs in a dimeric nanocluster (left) and corresponding
electron transmission intensity profile along the blue line (right). The average of the maximum and minimum electron
transmission intensities is calculated from the intensity profile for each AuNP in a cluster. The 50% value is used to
estimate the gap edge for both AuNPs and to measure the actual gap size. Scale bar: 20 nm.

Figure 3.15 TEM image of silver nanoclusters formed after co-incubation of 40 nm sGFP-AgNPs and 40 nm M3-
AgNPs. Scale bar: 200 nm. Inset: nanogap size measurement of a dimeric AgNP cluster. Scale bar: 20 nm.

GFP chromophore activation within plasmonic hot spots in metal NP clusters
Beyond inducing clustering, the reconstruction of FPs at the interface between AuNPs is
expected to additionally trigger the formation of plasmonic hot spots by increasing near-field
plasmon coupling and SERS enhancements at each nanogaps [253]. Within nanogaps, the re-
125

assembly of FP fragments is also expected to activate the maturation of the GFP chromophore.
Indeed, once GFP is complemented, a tripeptide chromophore is rapidly formed by autocatalytic
cyclization of three key amino acids residues within its β-barrel scaffold [111, 175]. This cyclized
and matured chromophore displays peculiar Raman fingerprints in the 1500-1650 cm
-1
spectral
region [112, 113]. These include the C=C stretching mode of the exocyclic double bond at 1630
cm
-1
and a normal vibrational mode delocalized over the imidazolinone ring and exocyclic C=C
bond with bands at 1560 cm
-1
(neutral) and 1530 cm
-1
(anionic) depending on the ground-state
protonation of the chromophore itself [112, 113].

Figure 3.16 Raman spectra of flGFP (310 μM), flYFP (232 μM) and flCFP (363 μM) in aqueous buffer. For all three
FP variants, the three main chromophore fingerprints at 1530 cm
-1
, 1560 cm
-1
and 1660 cm
-1
are detected within ±10
cm
-1
of their expected spectral positions, together with weaker vibrational bands of the chromophores and additional
Raman bands attributed to the rest of the proteins. Detected Raman bands were assigned in reference to previously
corresponding Raman shifts [112, 113, 254, 255]. λex: 785 nm, Pex: 3.33 mW/µm2, acquisition time: 150 seconds.

In Raman spectra of highly concentrated solutions of flGFP, flYFP and flCFP, these three
main chromophore fingerprints were detected within ±10 cm
-1
of their expected spectral positions
(Figure 3.16). For the sGFP fragment, which carries the three pre-cyclized amino acid residues, a
detection of these vibrational signatures is not expected because the lack of 11th β-sheet
126

precludes maturation of the chromophore. To establish that the Raman fingerprints of the
tripeptide chromophore are indeed absent in sGFP but can be effectively activated upon
complementation with M3 peptides, we used silver island plasmonic substrates and compared the
SERS spectra of the flGFP with that of the sGFP fragment before and after complementation with
M3 peptides. The peculiar vibrational modes of the mature GFP chromophore were detected at
1527 cm
-1
, 1563 cm
-1
and 1633 cm
-1
for the flGFP and the M3-complemented sGFP but were
absent for the non-complemented sGFP alone (Figure 3.17a) which confirmed the lack of
chromophore cyclization in sGFP and established the possibility to activate its SERS signatures
upon complementation with M3 peptides. Similar chromophore vibrational bands were detected
when M3-AgNPs pre-incubated with sGFP fragments were added to silver plasmonic substrates,
an indication that the chromophore activation is also efficient at the surface of metal colloids
(Figure 3.18).

Figure 3.17 SERS spectra of full-length GFP, sGFP fragment, and complemented sGFP on silver islands plasmonic
substrates and colloidal AuNP clusters at different pH. (a) SERS spectra of full-length GFP (blue), M3 peptide-
complemented split-GFP (red), and non-complemented sGFP fragment (green) on 5 nm silver island substrates. (b)
Liquid SERS spectra of assembled AuNP clusters at pH 8.0 (left) and pH 6.0 (right).

127

Figure 3.18 SERS spectra of M3-coated silver nanoparticles added on 5 nm silver island plasmonic substrates shortly
after co-incubation with the sGFP fragment. The typical GFP chromophore fingerprints are observed at 1517 cm
-1
,
1555 cm
-1
and 1645 cm
-1
. *Assigned to the C-C-H in-plane deformation mode of the chromophore’s phenol ring or the
CH2/CH3 deformation mode of amino acids in complemented sGFP. λex: 532 nm, Pex: 140 µW/µm
2
, acquisition time:
30 s.

To assess if a similar SERS activation of the GFP chromophore fingerprints takes place
within the nanogaps of assembled NPs, clusters were formed using 40 nm M3-AuNPs and 40 nm
sGFP-AuNPs, a size of NP that provides large near-field enhancements for SERS [253]. Raman
spectra of the flGFP and SERS spectra of M3-AuNPs, sGFP-AuNPs or the assembled AuNP
clusters were obtained in buffers at pH 8.0 or 6.0 to promote either the anionic or the neutral form
of the chromophore, respectively. Spectra were obtained at the non-resonance excitation
wavelength of 785 nm to avoid interference from fluorescence during Raman of flGFP and to
estimate accurately SERS enhancement factors. For the flGFP, which encodes the EGFP-like
mutations F64L and S65T and has a reported pKa of 5.5 [175], the Raman signature of the
anionic chromophore imidazolinone/exocyclic C=C mode at pH 8.0 was observed at 1538 cm
-1

and that of its neutral form at pH 6.0 was detected at 1549 cm
-1
, as previously reported for EGFP
(Figure 3.19) [256]. For AuNP clusters assembled for 12 hours, the SERS signature of the anionic
chromophore was detected at 1527 cm
-1
after deconvolution of vibrational modes from the
surrounding pH 8.0 buffer (Figure 3.17b). No GFP SERS signal was detected for colloidal
128

solutions of sGFP-AuNPs and M3-AuNPs at the same reaction time (Figure 3.19), an indication
that the chromophore SERS fingerprint arises only when stable AuNP clusters are formed. SERS
spectra of the AuNP clusters acquired at pH 6.0 displayed a shift of the 1527 cm
-1

imidazolinone/exocyclic C=C mode toward 1567 cm
-1
(Figure 3.17b) as expected when the
neutral form of the chromophore is favored over its anionic form at this lower pH [112, 256].
Similar pH-dependent shifts of the GFP chromophore Raman signature were also observed for
AgNP nanoclusters (Figure 3.19), confirming that the detected SERS signal result from the
assembly of the split FP fragments into a folded and mature GFP at the interface between NPs.

129

Figure 3.19 SERS spectra of individual and clustered NPs in buffer solutions. a) SERS and Raman spectra of sGFP-
AuNPs (top), M3-AuNPs (middle) and NaPT buffer (bottom) at pH 8.0. In the buffer alone and in the absence of AuNP
clusters, no GFP chromophore vibrational signature is detected. λex: 785 nm, Pex: 20 mW/µm
2
, acquisition time: 30 s. b)
Raman spectra of flGFP at pH 8.0 in TNG buffer (top left) and at pH 6.0 in acetate buffer (top right) and corresponding
Raman spectra of pH 8.0 TNG buffer alone (bottom left) and pH 6.0 acetate buffer alone (bottom right). The 1538 cm
-1

imidazolinone/exocyclic C=C mode of the anionic GFP chromophore in flGFP is detected among vibrational modes
from the TNG buffer at pH 8.0. Change in pH induces the neutral form of the chromophore and a shift of the 1538 cm
-1

mode toward 1549 cm
-1
. λex: 785 nm, Pex: 20 mW/µm
2
, acquisition time: 30 s. c) SERS spectra of assembled 40 nm
M3-AgNP and sGFP-AgNP clusters at pH 8.0 (left) and pH 6.0 (center) and Raman spectra of NaPT buffer at pH 6.0
(right) obtained for an excitation at 532 nm to optimize SERS responses. The anionic imidazolinone/exocyclic C=C
mode was detected at 1545 cm
-1
in the SERS spectrum of AgNP clusters at pH 8.0. The neutral mode was detected at
1570 cm
-1
in SERS spectra at pH 6.0. λex: 532 nm, Pex: 17 mW/µm
2
, acquisition time: 30 s.

When accounting for the concentration of AuNPs and the cluster size distribution at 12
hours, the experimental SERS enhancement factor of the chromophore 1527 cm
-1
mode at pH 8.0
with off-resonance excitation at 785 nm was 1.07x10
6
fold compared to flGFP. For comparison,
the theoretical SERS enhancement factors of this off-resonance vibrational mode calculated by
finite-difference time-domain modeling for a similar size distribution of AuNP clusters with GFP-
seeded plasmonic nanogaps of 2 nm or 4 nm were 2.39x10
6
fold and 3.86x10
4
fold respectively;
which correspond to a 2-fold and a nearly 30-fold difference with the experimental enhancement
(Figure 3.20). Off-resonance conditions lead to a decrease by 10
1
-10
2
in SERS enhancement
factor [221], which is expected to be in the range of 10
8
-10
9
under on-resonance conditions with 2
nm gap size [253]. The total SERS enhancement factor per cluster is calculated by summing the
integrated SERS enhancements within a 2x2x2 nm volume centered at the GFP position over
each nanogap within each cluster (e.g. 1 nanogap for dimers, 2 nanogaps for trimers, 3 nanogaps
for tetramers, etc.). The calculations are performed using:

𝐸𝐹
𝑆𝐸𝑅𝑆 −𝐺𝐹𝑃 ≈ ∑ |
E(𝜔 𝑒𝑥𝑐 )
E
0
(𝜔 𝑒𝑥𝑐 )
|
2
|
E(𝜔 𝑣𝑖𝑏 −𝐺𝐹𝑃 )
E
0
(𝜔 𝑣𝑖𝑏 −𝐺𝐹𝑃 )
|
2

3.1

where, for each nanogap, E is the amplitude of the local maximum electric field, E 0 is the
amplitude of the input source field polarized parallel to the long axis of each cluster, here 1 V/m,
E(ω exc)/E 0(ω exc) is the enhanced field at 785 nm laser excitation and E(ω vib-GFP)/E 0(ω vib-GFP) is
130

the enhanced field at the 1530 cm
-1
Stokes-shifted wavelength of the GFP chromophore
imidazolinone/exocyclic C=C mode, here 892 nm. For more information about the modeling and
the calculation method, see Chung et al. [253].
To estimate the theoretical SERS enhancement factor from solutions of AuNP clusters,
the cluster size distribution from Figure 3.13 was taken into account (dimers 28%, trimers 13%,
tetramers 7%, pentamers 3 % and hexamer 2%) and the total SERS enhancement for each type of
cluster weighted by these percentages was summed. The SERS enhancement from non-clustered
AuNPs (47%) was omitted, as it is comparatively negligible. For solutions with 2 nm gap AuNP
clusters the theoretical SERS enhancement is: 2.39x10
6
fold. For solutions with 4 nm gap AuNP
clusters the theoretical SERS enhancement is: 3.86x10
4
fold. These represent upper limit
electromagnetic enhancement factors as they are calculated for idealized and perfectly ordered
clusters.
The good agreement between measured and theoretical SERS enhancements for the 2 nm
nanogaps indicates that the observed SERS fingerprint at 1527 cm
-1
stems from the activation of
the GFP chromophore within uniform and 2 nm-sized plasmonic hot spots, as initially implied by
the measured nanogap dimensions (Figure 3.8d). Importantly, these data show that the split FP
guided self-assembly of AuNP clusters results in the formation of stable and well-defined
plasmonic hot spots that are homogenously seeded by a precise positioning and transverse
orientation of the complemented GFP within each nanogap. The absence of other protein or
peptide Raman bands in the solution SERS spectra is additional evidence of the controlled
formation of uniform hot spots within stable nanoclusters.

131

Figure 3.20 Finite-difference time-domain calculated theoretical SERS enhancement factors of the GFP
imidazolinone/exocyclic C=C mode at 1530 cm
-1
for different sizes of AuNP clusters with GFP-seeded nanogap
dimensions of 2 or 4 nm. The clusters are modeled as linear chain assemblies of 40 nm AuNPs with n = 2, 3, 4, 5 or 6
AuNPs per chain.

Indeed, such modes were only detected when aggregation of the NPs was induced (Figure
3.21). The large experimental SERS enhancement factor obtained at non-resonant 785 nm
excitation also demonstrates that the activated GFP chromophore within AuNP clusters can be
detected with high sensitivity. In these clusters, the concentration of complemented GFP was
estimated at 0.6 nM, which implies that our ensemble SERS measurements in solution are
performed in a single molecule detection regime. For resonant excitation of the AuNPs clusters at
633 nm, the chromophore imidazolinone/exocyclic C=C mode at 1530 cm
-1
is expected to be
further amplified with SERS enhancement factors in the range of 10
8
-10
9
fold [253], a level of
enhancement generally sufficient for single molecule SERS. Thus, in addition to serving as
molecular glue to assemble AuNPs clusters with well-defined nanogaps, the complemented FP
fragments also trigger the self-activation of the GFP chromophore within uniform and small
132

plasmonic hot spots. The activated chromophore effectively acts as a highly sensitive and
activatable SERS reporter of the clustering process.

Figure 3.21 SERS spectra of sGFP-AuNPs, M3-AuNPs, and AuNP clusters assembled for 12 hours from a 50 μl drop
deposited on a SiO2 wafer. The spectra were taken after aggregations of the colloidal solutions following the
evaporation created by the laser. AmIII bands and CH2/CH3 scissor bending from sGFP and M3 peptides were
observed from all samples because of the artificial clustering. However, GFP chromophore Raman fingerprints, which
require binding and maturation processes to form, were only observed from the complex sample. λex: 785 nm, Pex: 33
mW/µm
2
, acquisition time: 60 s [112, 113, 254, 255].

Site-directed assembly of AuNP clusters in live cells
To test whether AuNP clusters could be assembled and activated via biomarker-assisted
clustering directly in live cells, we used HeLa and U2OS cells expressing extracellular
transmembrane and GPI-anchored avidin-fusions [257] and targeted these surface biomarkers
with 40 nm M3-AuNPs and 40 nm sGFP-AuNPs via their surface-attached biotin-PEG 600
moieties (Figure 3.22a). Both sets of biotinylated NPs rapidly recognized the avidin-fusions and
bound specifically to expressing cells (Figure 3.23). The membrane-bound AuNPs were imaged
by total internal reflection fluorescence (TIRF) microscopy by exploiting the non-linear
luminescence of spherical AuNPs near borosilicate glass coverslips [258].
133

134

Figure 3.22 Cell targeting and plasma membrane clustering of biotinylated AuNPs. (a) Schematic of biotinylated M3-
AuNPs and sGFP-AuNPs targeted to avidin fusion proteins at the plasma membrane of cells and their assembly into
SERS active clusters. Not to scale. (b) Bright field (left), single frame TIRF microscopy image (middle) and maximum
intensity projection TIRF image from multiple frames (∑Imax, right) of biotinylated M3-AuNPs bound to GPI-avidin
fusion proteins and diffusing at the plasma membrane of live HeLa cells. Scale bar: 7 μm. c) Scanning electron
microscope images of U2OS cells co-expressing the transmembrane and the GPI-anchored avidin fusion proteins and
targeted with biotinylated sGFP-AuNPs alone (top panel), biotinylated M3-AuNPs alone (bottom panel) or both
biotinylated sGFP-AuNPs and M3-AuNPs simultaneously (middle panel). White arrows point towards endocytic
membrane structures, blue arrowheads point towards AuNPs monomers and red arrowheads point towards some of the
AuNP nanoclusters presented in insets. The plus and minus signs identify the avidin-expressing and non-expressing
cells, respectively. Scale bars: 2 μm (left panels), 200 nm (insets of left panels), 1 μm (right panels) and 100 nm (insets
of right panels). (d) Cluster size distributions of AuNPs on targeted U2OS cells.

When targeted separately to the avidin-fusions and imaged by TIRF, individual M3-
AuNPs and sGFP-AuNPs diffused rapidly in the plane of the plasma membrane and over the
entire surface of expressing cells (Figure 3.22b), consistent with the expected rapid lateral
mobility of transmembrane and GPI-anchored proteins in cells [168, 257]. This indicated that
after specific binding to plasma membrane biomarkers, AuNPs remain monodispersed and high
mobile, which is critical for their self-assembly into SERS nanoclusters on cells.

Figure 3.23 Reactivity of biotinylated AuNPs with streptavidin and specific targeting at the plasma membrane of live
cells expressing extracellular avidin fusion proteins. a) 0.8% agarose gel shift assay after incubation of sGFP-AuNPs or
M3-AuNPs with decreasing concentrations of streptavidin (3.33 μM, 0.83 μM, 0.42 μM, 0.17 μM, 41.67 nM, 25 nM,
16.67 nM, 3.33 nM) for 45 min at room temperature. While no shift is observed for non-biotinylated M3-AuNPs and
135

sGFP-AuNPs, discrete shifts are seen for the biotinylated NPs at high streptavidin concentrations. This indicates that
surface attached biotin-PEG moieties effectively react with streptavidin despite the additional presence of M3 peptide
or sGFP fragments. b) Top: Fluorescence wide-field microscopy image of fixed HeLa cells expressing the membrane
avidin fusions and labeled with biotin-Alexa594 fluorophore (yellow). Middle: Dark field microscopy image of the
cells (cyan). Bottom: Overlay of fluorescence and dark field images. Expressing cell targeted with biotinylated sGFP-
AuNPs display larger dark field signals than non-targeted cells, which only show weak dark-field signals due to
scattered light from membranes and internal organelles. Scale bars 20μm.

When M3-AuNPs and sGFP-AuNPs were simultaneous targeted on cells co-expressing
both the transmembrane and the GPI-anchored avidin fusions, a variety of clusters quickly
formed at the cell surface. Within an hour of incubation, AuNP dimers, trimers, tetramers and
longer AuNPs chains were found at the plasma membrane, and 40% of all the membrane-bound
AuNPs were clustered (Figure 3.22c, d and Table 3.1). In comparison, separately targeted M3-
AuNPs or sGFP-AuNPs showed minimal clustering on cells, with 91% and 87% of the NPs
remaining as monomers, respectively (Figure 3.22c, d and Table 3.1). Together with these
discrete AuNPs clusters, a few large and circular aggregates were present at the plasma
membrane (insets of Figure 3.22c). These larger AuNPs aggregates, which likely represent early
stage endocytic events at the cell surface, were also detected on cells targeted with only M3-
AuNPs (insets of Figure 3.22c).

Table 3.1 Frequency of cluster size distribution for sGFP-AuNPs and M3-AuNPs functionalized with biotin-PEG600 or
non-biotinylated-PEG600 and simultaneously co-targeted or separately targeted on U2OS cells. Cells co-express both
the GPI- and transmembrane avidin fusions.

When cells were targeted with either M3-AuNPs or sGFP-AuNPs lacking biotin-PEG 600
moieties, 90 % of the membrane-bound AuNPs remained monomeric (Figure 3.22d, Figure 3.24
136

and Table 3.1), an indication that the plasma membrane clusters were not produced by the pre-
assembly of AuNPs in the cell media before binding to cells.

Figure 3.24 SEM images of U2OS cells expressing both avidin fusions and co-targeted with AuNPs lacking surface
biotin. a) Co-targeting of M3-AuNPs without biotin together with sGFP-AuNP-biotin-PEG600. b) Co-targeting of
sGFP-AuNPs without biotin together with M3-AuNP-biotin-PEG600. While specific labeling of expressing cells (+) is
achieved, no plasma membrane clustering of nanoparticle is observed if one of the two AuNPs lacks biotin-PEG600.
Blue arrowheads point towards individual AuNPs presented in the insets. Scale bars: left panels: 2 μm, right panels: 1
μm, insets: 100 nm.

Replacing biotin-PEG 600 by biotin-PEG 2000 on AuNPs had no significant influence on the
clustering efficiency on cells (Figure 3.25).
137

Figure 3.25 SEM images of U2OS cells expressing both avidin fusions and incubated with sGFP-AuNP-biotin-PEG2000
and M3-AuNP-biotin-PEG2000 independently or together. As with AuNPs modified with biotin-PEG600 efficient
clustering is observed at the cell plasma membrane of expressing cells (+) for co-targeted nanoparticles only. Blue and
red arrowheads point, respectively, towards individual or clustered AuNPs presented in the insets. Scale bars: Left
panels: 2 μm, right panels:1 μm, insets: 100 nm.

However, when biotinylated AuNPs were co-incubated with cells expressing only the
GPI-avidin fusion, clusters were mainly dimeric; suggesting that the type of membrane anchorage
for the targeted biomarkers can influence the clustering process (Figure 3.26).
138

Figure 3.26 SEM images of HeLa cells expressing only the GPI-avidin fusion and incubated with sGFP-AuNP-biotin-
PEG2000 and M3-AuNP-biotin-PEG2000 independently (top) or together (bottom). Clusters are primary dimeric (red
arrowheads and insets) and residual monomeric AuNPs are observed at the plasma membrane (white arrows). Scale
bars: 1μm; insets: 100 nm.

The clustering of AuNPs at the cell membrane is therefore driven by the bimolecular
complementation of the split FP fragments appended to their surface and is not the result of
random aggregations or accumulation in membrane pits or cavities. The kinetic of clustering at
the plasma membrane is much faster than in solution and is primarily assisted by the lateral
diffusion of the targeted biomarkers. This is consistent with the fact that, for binding reactions
that also depends on diffusion, the reaction efficiency is increased when the number of
dimensions in which diffusion occurs is reduced. Indeed, the mobility of AuNPs and their
diffusion within the two-dimensional plane of the cell membrane facilitate molecular interactions
and steric fit between the surface-appended FP fragments and improve the clustering efficiency
compared to assemblies in solution where AuNPs undergo three-dimensional and rotational
diffusions (Figure 3.27).
139

Figure 3.27 Schematic of AuNP diffusion difference in NaPT buffer (top) and at the cell plasma membrane (bottom).
Cartesian coordinate axes represent both translational and rotational diffusion in three-dimension inside the buffer
system, whereas it represents only translational diffusion in two-dimension at the cell plasma membrane.

SERS microscopy imaging of metal NP clusters in targeted cells
Cells grown at confluency and transfected with both avidin biomarkers were co-targeted
for 1 hour with 40 nm M3-AuNPs and sGFP-AuNPs and imaged live or after chemical fixation
by SERS microscopy. Wide-field SERS imaging of fixed cells was performed in PBS buffer at
pH 8.0 using a 532 nm excitation and a Raman hyperspectral imager based on Bragg tunable
filters [150]. In the absence of AuNPs, no GFP chromophore fingerprints were detected in SERS
spectra and only a diffuse background cell autofluorescence was observed when spectral images
were reconstructed at 1527 cm
-1
to localize the imidazolinone/exocyclic C=C Raman mode
(Figure 3.28a). Likewise, no SERS fingerprints of the GFP chromophore were detected for cells
targeted with only M3-AuNPs or only sGFP-AuNPs and spectral images at 1527 cm
-1
displayed
the cell background and a few low intensity spots across the plasma membrane (Figure 3.28a).
These small increases in local background were attributed to weak photoluminescence signal
140

contributions from individual AuNPs when excited at 532 nm [259]. In contrast, the typical
vibrational signatures of the GFP chromophore were clearly visible in SERS spectra taken at
various positions along the plasma membrane of cells co-targeted by both M3-AuNPs and sGFP-
AuNPs (Figure 3.28a), a direct confirmation that the biomarker-assisted clustering of AuNPs
effectively induces the activation of the GFP chromophore within plasmonic hot spots at the cell
surface. In these spectra, the three chromophore fingerprints were detected within ±10 cm
-1
of
their expected position, with the 1527 cm
-1
imidazolinone/exocyclic C=C mode often dominating
the signal (Figure 3.28a). When images were reconstructed at this wavenumber, targeted cells
were specifically distinguished from non-transfected cells that did not express the biomarkers
(Figure 3.28a).
141

142

Figure 3.28 Targeted SERS imaging of cells with split-FP assembled metal nanoclusters. a) SERS microscopy images
of fixed cells at the GFP chromophore 1527 cm
-1
imidazolinone/exocyclic C=C Raman mode and corresponding SERS
spectra on cells expressing the avidin biomarkers and targeted by biotinylated M3-AuNPs and sGFP-AuNPs separately
or simultaneously. Colored arrows in images point toward single pixels whose SERS spectra are represented in
matching colors. The three typical GFP chromophore vibrational modes are indicated by dash lines on spectra. b) SERS
microscopy image of live cells co-targeted by biotinylated M3-AuNPs and sGFP-AuNPs and reconstructed at a
1535±15 cm
-1
spectral window. The deconvolved SERS spectrum corresponds to one pixel in the cell image as
indicated by the arrow. c) SERS image of live cells co-targeted by biotinylated M3-AgNPs and sGFP-AgNPs and
reconstructed at a 1550±15 cm
-1
spectral window. The SERS spectrum corresponds to the individual pixel indicated by
the arrow in the cell image. d) SERS image of live cells in hypotonic buffer after co-targeting of biotinylated M3-
AuNPs and sGFP-AuNPs and reconstruction at a 1550±15 cm
-1
. The SERS spectrum corresponds to one pixel in the
cell image as indicated by the arrow. All scale bars: 10 µm.

Although the SERS signal at 1527 cm
-1
was detected across the cell surface, its intensity
was not homogenous. While large areas of the plasma membrane and smaller punctuates
displayed high SERS signals, other membrane domains lacked the spectral fingerprints of GFP
and show background signals similar to those observed in cells targeted with only M3-AuNPs or
sGFP-AuNPs (Figure 3.28a). This disparity in SERS signal distribution is consistent with the size
distribution of AuNP clusters at the cell plasma membrane and the presence of residual non-
clustered AuNPs (Figure 3.22c). Indeed, because the number of activated GFPs and SERS
enhancement factors both scale with the size of clusters [253] (Figure 3.20), different AuNP
clusters may display different GFP SERS intensities. In addition, not all complemented GFP
within clusters might be matured after 1 hour clustering on cells because the maturation half-time
of the chromophore after complementation of the FP fragments is about 30 min. Together, these
factors participate to variations in GFP SERS signal intensity across targeted cells.
For live cell assays, cells were incubated with 40 nm M3-AuNPs and sGFP-AuNPs for
one hour, washed and further imaged over a one-hour period in pH 8.0 PBS buffer by confocal
Raman microscopy at 532 nm excitation. As for fixed cells, the imidazolinone/exocyclic C=C
modes of the GFP chromophore were detected within ±10 cm
-1
of their expected positions, at
about 1524 cm
-1
and 1551 cm
-1
in SERS spectra of targeted cells (Figure 3.28b). In images
reconstructed at 1535±15 cm
-1
to encompass the Raman signals from both fingerprints, the SERS
143

signal was diffused at the cell surface and SERS spectra from individual pixels were noisier than
in fixed cells (Figure 3.28b), consistent with the lateral diffusion of the targeted avidin
biomarkers and the expected mobility of the assembled AuNP clusters in live cells. Similar
clustering-dependent activation of the GFP chromophore Raman fingerprints were obtained when
40 nm M3-AgNPs and sGFP-AgNPs were co-targeted to cells and imaged under the same
conditions (Figure 3.28c), confirming that different types of metal NPs can self-assemble into hot
spot SERS nanoprobes directly in live cells via in situ complementation of split FP fragments.
With AgNP clusters, SERS signal-to-noise detections of the GFP chromophore fingerprints were
significantly improved compare to AuNPs (Figure 3.28c), because plasmon-plasmon coupling
between AgNPs red-shifts their maximum near-field enhancement wavelength [235, 260] in
resonance with the 532 nm excitation used for cell imaging. To confirm that the detected 1530-
1560 cm
-1
SERS vibrational bands arise from the complementation of GFP, we also stimulated
the early endocytosis of 40 nm M3-AuNPs and sGFP-AuNPs clusters by imaging targeted cells in
a hypotonic PBS buffer [261]. Under this condition, the neutral GFP chromophore Raman
signature at about 1560 cm
-1
was primarily detected in SERS spectra, consistent with an
accumulation of the nanoclusters into early endosomal compartments which have a slightly acidic
pH of 6.0-6.5 [262] (Figure 3.28d). Complementary metal NPs functionalized with split FP
fragments can therefore rapidly self-assemble into active hot spot SERS nanocluster probes when
co-targeted to diffusing plasma membrane biomarkers in live cells and the in-situ activation of the
GFP chromophore Raman fingerprints within NP clusters allows highly specific and single cell
SERS imaging.

144

Photoacoustic imaging of AuNP clusters on targeted cells
In addition to providing versatile plasmonic platforms for SERS, metal NPs are also good
exogenous contrast agents for photoacoustic detection of targeted cells and tissues [148, 263]
where optical excitations induce transient thermal expansions around NPs and generate acoustic
pressure waves detectable by ultrasound imaging (Figure 3.29) [133, 264]. Indeed, once targeted
to cells, the self-assembly of 40 nm M3-AuNPs and sGFP-AuNPs into nanoclusters resulted in
strong optoacoustic signal amplifications and individual cells expressing the avidin biomarkers
could be specifically imaged by photoacoustic microscopy with pulsed laser excitation at 532 nm
(Figure 3.30a). On targeted cells, the photoacoustic signal amplitude of the assembled AuNP
clusters was about twice that of cells labeled with sGFP-AuNPs only (Figure 3.30b), consistent
with the fact that clustered AuNPs provide enhanced photoacoustic signals compared to
individual AuNPs [265] because thermal interfacial conductivity increases significantly within
closely spaced AuNPs, notably along NP chains [266]. For both clustered and non-clustered
AuNPs, the photoacoustic signals scaled linearly with the range of laser excitation energy tested
(Figure 3.30c) an indication that the enhanced photoacoustic signal amplitude stems from the
formation of discrete M3-AuNPs and sGFP-AuNPs clusters on targeted cells rather than from
endocytosed AuNP aggregates which often display non-linear photoacoustic responses with
increasing excitation intensities [267]. Thus, in addition to allowing activatable SERS detections,
the in-situ and split FP-assisted clustering of AuNPs yields strong photoacoustic signal
amplifications compared to individual AuNPs on targeted cells. The self-assembled metal
nanoclusters effectively serve as bimodal contrast agents for SERS and photoacoustic microscopy
imaging of individual cells.
145

Figure 3.29 Schematic of the laser-scanning photoacoustic microscopy system [148].

Figure 3.30 . Photoacoustic imaging of in-situ assembled split-FP AuNP clusters on cells. a) Photoacoustic microscopy
images of individual U2OS cells among a 100% confluent field after targeted clustering of biotinylated M3-AuNPs and
sGFP-AuNPs on cells that transiently express plasma membrane avidin biomarkers. Scale bars: 50 µm (left) and 20 µm
(right). b) Photoacoustic signal amplitudes from similar fields of cells targeted with both M3-AuNPs and sGFP-AuNPs
(split-FP clustered) or with sGFP-AuNPs only (non-clustered). ***: p<0.01, T-test. c) Photoacoustic signal amplitudes
from targeted cells at increasing laser excitation energy. Lines represent linear regression fit of the data.
146

Conclusion
We have shown that different split-FP fragments can be used to control the assembly of a
variety of plasmonic NPs into activatable SERS nanocluster probes with well-defined hot spots in
solution and directly in live cells. The irreversible reconstruction of fully folded FPs from
complementary split FP domains appended to the surface of metal NPs provides stable
nanoclusters having interfacial nanogap sizes determined by the precise positioning and
transverse orientation of the FP β-barrel structure. While other high-affinity protein domains
might be employed for similar NP assemblies, the autocatalytic formation of the FP chromophore
provides a direct read out of the clustering process and serves as a highly sensitive Raman
reporter for low background SERS imaging in complex biological environments with, potentially,
single molecule SERS detection sensitivity. As such, split FPs represent a versatile family of
biocompatible scaffolds and Raman reporters, which can be engineered by site-directed
mutagenesis to generate additional FP variants and photo-controllable FPs [268, 269] for spectral
matching with the near-field optical spectra of metal NP clusters and for advanced Raman
scattering applications such as resonance and photo-switchable SERS imaging.
Upon targeting to diffusing cell surface biomarkers, complementary AuNPs rapidly
organized into linear clusters. This linear ordering of NPs at the plasma membrane together with
the 2 nm gap size formed by the complemented FPs provides large SERS amplifications of the FP
chromophore for high sensitivity and single cell SERS imaging and induce strong signal
amplifications for photoacoustic imaging. Although we primarily used off-resonance 532 nm
excitation of the AuNP clusters in cells to directly compare their photonic responses with that of
individual NPs, plasmon-plasmon couplings within 40 nm AuNP linear clusters induce a large
shift of their maximum near-field enhancement wavelength towards 630 nm [253]. Imaging
AuNPs clusters at this resonant excitation wavelength is expected to provide SERS detection of
147

the GFP chromophore vibrational modes at the onset of the near-infrared spectrum with two
orders of magnitude higher SERS enhancement factors [253] and to further increase signal
amplitudes for photoacoustic imaging of targeted cells. Applying this split FP assembly strategy
to other plasmonic nanomaterials such as nanoshells [270] might provide even larger SERS
enhancement factors for targeted Raman imaging and photoacoustic excitation of cells and tissues
further in the near-infrared.
While we used non-endogenous plasma membrane biomarkers to demonstrate specific
clustering in live cells, other endogenous surface markers, such as those overexpressed on cancer
cells, can be targeted using complementary AuNPs functionalized with antibodies. Compared to
pre-activated SERS probes, this dual NP assembly and in situ SERS activation strategy should
provide highly selective detection of pathogenic cells over normal cells, notably if
complementary NPs are targeted to two different overexpressed biomarkers. Indeed, because
most cancer markers are also expressed on normal cells but are much lower levels, smart SERS
probes that switch on as a function of biomarker molecular density and diffusion, as shown here,
can presumably reduce false positive detections, notably for in vivo imaging, where non-specific
binding and uptake of nanoprobes by tissues are significant. We note that a good knowledge of
the biomarker membrane mobility and potential lipid phase separation upon clustering is
important to ensure effective self-assembly of NP clusters on targeted cells. Further active control
of the nanocluster assembly and SERS signals might also be achieved by caging synthetic M3
peptides, for instance with photo-uncageable chemical moieties or with protease-responsive
peptide sequences for uncaging and clustering within specific tissues in vivo.
Overall, the bottom-up assembly of colloidal metal NPs using split FP scaffolds as both
molecular glue and Raman reporters addresses the long-standing issue of forming nanostructures
having well-defined Raman hot-spots and that of site-specific SERS activation of nanoprobes in
148

complex biological milieus. It provides a novel approach to remotely assemble nanocluster
probes on biological targets for multimodal SERS and photoacoustic microscopy imaging with
high sensitivity and selectivity, both in cells and in vivo.

Material and Methods
Expression and purification of sGFP
Plasmids encoding sGFP with a N-terminal 6xHis-tag, a GSS linker sequence, a thrombin
cleavage site, a tetracysteine motif and a flexible GGSGG linker domain (Figure 3.2) were
transformed in a BL21(DE3) E. coli strain for protein expression. 50 ml LB overnight starter
culture (10 μg/ml kanamycin) was prepared with a transformed E. coli colony. 25 ml of the
overnight culture was inoculated into 1 L LB (35 μg/ml kanamycin) and the culture was
incubated in a shaker at 37 °C until OD 600 reaches ~0.6. The culture was cooled down at room
temperature for 20 min. After 1 mM IPTG induction, the culture was incubated overnight at 20
°C. Cells were harvested at 4000 g, for 30 min, at 4 °C. The cell pellet was washed with 20 ml ice
cold PBS at 4000g, for 30 min, at 4 °C. The cell pellet was suspended in TNG/imidazole buffer
(100 mM Tris-HCl, 150 mM NaCl, 10 % glycerol, 10 mM imidazole, pH 8.0). 1x HALT protease
inhibitor, 0.5 mM TCEP, 5 μl benzonase nuclease/g cell pellet, and 5 ml 1x bugbuster/g cell
pellet were added and incubated for 30 min at room temperature for cell lysis. The sample was
centrifuged at 16000 g, for 15 min, at 4 °C and the supernatant was collected. Since sGFP is
expressed with a 6xHistag, Ni-resin beads were used to purify the sGFPs. TNG buffer with 10
mM imidazole and 150 mM imidazole were used as wash buffer and elution buffer, respectively.
The sample was dialyzed against 1 L TN buffer (100 mM Tris-HCl, 150 mM NaCl, pH 8.0) for 1
hour at 4 °C to remove imidazole. TN buffer was replaced and dialysis was continued overnight
149

at 4 °C. A BCA protein assay was used to define the concentrations of the split-GFPs. Thrombin
(15 U/mg protein) was used to remove 6xHistag from the proteins. Thrombin cleavage was
performed for 20 min at room temperature in the presence of 1 mM TCEP. p-aminobenzamidine
beads were used to eliminate the thrombin after cleavage. The sample was dialyzed against 1 L
TN buffer for 1 hour at 4 °C to remove TCEP. TN buffer was replaced and dialysis was continued
for overnight at 4 °C. 10 % glycerol was added to the dialyzed samples before freezing. The
proteins were frozen using liquid nitrogen and stored at -80 °C. Full-length super-folder GFP
(flGFP) and other split-FP variants were expressed and purified following this same protocol.

Characterization of recombinant sGFP
The purity of sGFP was assessed by SDS gel electrophoresis and by direct comparison
with a commercial sGFP (Sandia Biotech), before and after purification on Ni-resin beads and
cleavage of the 6xHis-tag by thrombin. The recombinant protein was detected at ~26 KDa and
was >90 % pure, with a small percentage of unreduced dimers (~9 %, Figure 3.3). After thrombin
cleavage, the presence and the activity of the sGFP N-terminal tetracysteine motif were assessed
by fluorescence labeling with ReAsh (Adams 2002). Gel electrophoresis was performed on 1 %
agarose gels after ReAsH labeling and/or complementation of sGFP with an excess of M3
peptides. Gels were scanned on a Biorad, Molecular Imager FX, with appropriate laser excitation
and emission filters for GFP, ReAsH and GFP-to-ReAsH FRET detections (Figure 3.4). ReAsH
binding and GFP-to-ReAsH FRET detection indicated that the tetracysteine motif is effectively
present and active at the N-terminus of sGFP and that its activity is not influenced by the binding
of complementary M3 fragments.

150

Surface functionalization of nanoparticles with sGFP and M3 fragments
To functionalize AuNPs with tetracysteine-sGFP, 0.5 μM sGFP was mixed with 300 μl
citrate capped AuNPs (40 or 10 nm in diameter, optical density of 1.0, Sigma) in the presence of
0.5 μM thiolated-PEG-biotin (NanoCS) in NaPT buffer (8 mM NaH2PO4, 50 mM NaCl, 0.05 %
tween-20 at pH 8.0) and incubated overnight at room temperature. Excess protein was removed
by multiple rounds of centrifugation at 7000 g for 10 minutes. sGFP-coated AuNPs were re-
suspended in NaPT buffer before use. In addition to the surface modification of AuNPs with
sGFP, sYFP and sCFP expressing the same N-terminal tetracysteine motif were used to
functionalize AuNPs. The expression, the purification and the coating procedure for these split FP
variants was the same as for sGFP. As observed for sGFP, the surface coating of AuNPs with
sYFP or sCFP resulted in stabilization of the colloidal nanoparticles which were monodispersed
and did not show signs of aggregation when analyzed by DLS or TEM (Figure 3.6 and Figure
3.7).
To coat AuNPs with the M3 peptide fragment, synthetic and cysteine-modified M3
peptides (C-acplinker-GSGGGSTSRDHMVLHEYVNAAGIT, Lifetein LLC, purity >75 %) were
used. 200 μM of M3 peptide and 20 μM cysteine-PEG-biotin were mixed with 300 μl AuNPs (40,
20 or 10 nm in diameter, optical density of 1.0, Sigma) in NaPT buffer. After overnight
incubation at room temperature, excess peptide was removed by multiple rounds of centrifugation
at 7000 g for 10 minutes. M3-coated AuNPs were resuspended in NaPT buffer before use.
To functionalize oleylamine-stabilized AuNPs or AgNPs (10 nm diameter), two
multicysteine synthetic peptides previously designed to bind to CdSe/ZnS quantum dots [168,
271] were used. These include (i) a small spacer-peptide with amino acid sequence
KGSESGGSESGFCCFCCFCCF that provides hydrophilicity and makes space for (ii) a M3
peptide sequence FCCFCCFCCFGGSESG-(dPEG6)-GSGGGSTSRDHMVLHEYVNAAGIT
151

that is more hydrophobic and provides reactivity to sGFP on AuNPs. Peptides dissolved in
DMSO were rapidly mixed with AuNPs or AgNPs in toluene, and a few microliters of
tetramethylammonium hydroxide (Sigma-Aldrich) was immediately added (i) to trigger the
formation of cysteine thiolates anion in the peptides, (ii) to remove the hydrophobic surfactant
from the surface of AuNPs or AgNPs and (iii) to drive the binding of the peptides on NPs, as
previously described2. A slurry pellet was obtained after a vigorous shaking step. The supernatant
was carefully removed and the pellet was dissolved in DMSO before a slow buffer exchange step
on a G-10 column (Harvard Apparatus) equilibrated with distilled water. The eluate was
extensively dialyzed against a NaP buffer in cellulose ester dialysis membranes (Spectra/Por®
Biotech, 100 kDa MWCO, Spectrum Laboratories, Inc.) to remove non-reacted peptides.

Assessing the presence of sGFP on AuNPs
To confirm the binding of sGFP at the surface of AuNPs we first compared the colloidal
stability of citrate-stabilized bare AuNPs with that of sGFP-coated AuNPs. AuNPs were run on
0.8% agarose gels with 1x TAE buffer pH 8. Under these conditions, citrate-stabilized AuNPs
rapidly aggregate and do not migrate in the gel, but AuNPs coated with sGFP or with flGFP are
stabilized against aggregation and migrate as narrow bands (Figure 3.5), indicating the effective
presence of the proteins at the surface [272]. In a more direct approach, we also detected the
presence of sGFP on AuNPs by immuno-blot assays against GFP directly on the nanoparticles
after protein coating and purification (Figure 3.1c). For blotting, a strip of transfer membrane was
first rinsed with methanol, water, and TBS buffer (20 mM Tris-HCl, 150 mM NaCl, pH 7.4).
AuNP samples were then spotted on a PVDF membrane (Biorad) and further incubated until they
were fully absorbed in the membrane but not dried out. The membrane was then rinsed with TBS
buffer and was incubated at room temperature for 30 min in blocking buffer, 5 % dry milk in
152

TBST (0.05 % tween-20 in TBS buffer). The blocking buffer was removed and the membrane
was washed 3 times for 5 min with TBST buffer. A polyclonal rabbit anti-GFP primary antibody
(ThermoFisher A-6455, 1:2000 dilution) was applied for 1 hour at room temperature in blocking
buffer. The membrane was washed 3 times in TBST buffer for 5 minutes. A goat anti-rabbit HRP
secondary antibody (ThermoFisher 32260, 1:4000 dilution) was then applied for 1 hour at room
temperature in blocking buffer. The membrane was further washed 3 times with TBST buffer for
5 minutes and SuperSignal West Pico chemiluminescent substrate (ThermoFisher, 34080) was
applied to the membrane. A Biorad Chemidoc system was used for chemiluminescence detection
with 10 min of exposure time. As seen in Figure 3.1c, bare AuNPs (citrate-stabilized) do not
induce an immuno-reaction when targeted by anti-GFP antibodies, but both flGFP and sGFP-
coated AuNPs do, which provides direct evidence of the stable anchoring of both proteins at the
surface of the metal nanoparticles.

Dynamic light scattering and transmission electron microscopy of AuNPs
The hydrodynamic diameter of AuNPs was measured on a dynamic light scattering
instrument (Wyatt Technology, DynaPro Titan) using 10 seconds acquisitions and a series of 30
repetitive acquisitions at 25 ˚C for each sample.
For electron microscopy, AuNPs were intentionally deposited at low density on TEM
grids to prevent the formation of drying-mediated 2D nanoparticle assemblies. 10 μl of AuNPs
diluted in NaPT were dropped on parafilm and TEM grids (Ted Pella, Inc. Carbon Type-B, 200
mesh, Copper) were inversed on the 10 μl samples for 20 minutes. After deposition, the TEM
grids were transferred on a drop of Milli-Q water to rinse off the buffer and avoid salt crystals. A
JEOL Jem-2100 (LaB6) microscope operated at 200 kV was used for imaging and a Gatan
software (GMS-3) was used for TEM image analyzes.
153

Formation of nanoclusters by assembly of split-fluorescent protein fragments and
competition assay
AuNP clusters were formed by co-incubation of equivalent amounts of sGFP-AuNPs
(3.9x10
11
NP/ml) and M3-AuNPs (3.9x10
11
NP/ml) for 12 hours in NaPT buffer at room
temperature. This co-incubation time resulted in the formation of different sizes of nanocluster
with more than 50 % of all AuNPs being clustered (see Figure 3.13). Following co-incubation,
AuNP samples were run at 120 V for 20 min in a 0.8 % agarose gel equilibrated with 1x TAE
buffer. To purify the nanoclusters, the shifted/smear bands corresponding to AuNP clusters were
cut and electro-eluted from the gel in cellulose ester dialysis membranes (Spectra/Por® Biotech,
100 kDa MWCO) in NaP buffer (8 mM NaH 2PO 4, 50 mM NaCl, pH 8.0). The purified
nanoclusters were collected from the dialysis membrane and stored at 4 °C before further use.
Alternatively, purification of some nanoclusters was done using multiple rounds of centrifugation
at centrifugal forces adapted to the size and the sedimentation velocities of the AuNPs used in the
assembly reaction. For instance, nanoclusters formed with 40 nm AuNPs reacted with 10 nm
AuNPs were purified by a few rounds of centrifugation at 5000 g for 5 min.
To compete with the assembly process, a large excess of free and non-cysteinilated M3
peptide (100 µM, RDHMVLHEYVNAAGIT, Lifetein LLC, purity >75 %) was added during the
co-incubation of sGFP-AuNPs and M3-AuNPs.

Statistical analyses of the size distribution of AuNP nanoclusters and measurement of
nanogap size between AuNPs
After 12 hours co-incubation of sGFP-AuNPs with M3-AuNPs, unpurified samples were
directly applied to TEM grids to perform statistical analyses on the formation of multimeric
nanoclusters and to assess the size heterogeneity of the clusters. A total of 405 AuNP monomers
154

and AuNP nanoclusters were evaluated from five independent experiments. In 53 % of cases,
AuNP nanoclusters with at least two AuNPs were observed. To define whether the formation of
AuNP nanoclusters was random or effectively driven by complementation between split-FP
fragments, we first compared the experimental distribution of AuNP nanocluster sizes with that
expected for a random clustering process. In case of random clustering, a Poisson distribution of
nanocluster sizes for a mean assembly efficiency of 53 % is expected (𝑃 (𝑘 ) =
𝜆 𝑘 𝑒 −𝜆 𝑘 !
with λ =
0.53). As shown in Figure 3.13, our experimental distribution is not well described by such a
Poisson distribution, indicating that the AuNP clustering process is not random.
To define which clustering kinetic regime leads to our observed distribution of cluster
sizes, we plotted our experimental size distribution of AuNP clusters after normalization and
log/log scale transformation as previously described [250]. In solution, the clustering of colloids
can be described by a diffusion-limited aggregation (DLA) model or a reaction-limited
aggregation (RLA) model [250, 251]. For irreversible binding between colloids these models give
markedly different aggregate’s dimension [250, 251] and very different cluster size distributions
[250, 251]. The DLA model leads (i) to a rapid formation of large, branched aggregates and (ii) to
cluster size distributions that are characterized by a peaked distribution [250]. The RLA model
leads (i) to the preponderant formation of smaller, more compact clusters due to the slower rates
of interaction and (ii) to cluster size distributions that are best described by a power law
distribution [250, 251]. As shown in Figure 3.8c, our nanocluster size distribution was very well
described by a power law fit, with a power coefficient of 1.7± 0.3, within the expect range for
RLA processes (range: 1.5-2) [250, 273, 274]. This indicates that the formation of the AuNP
nanoclusters is due to a reaction-limited aggregation as is expected for an assembly that is driven
by the irreversible bimolecular complementation between the two split-fluorescent protein
fragments.
155

The size of the nanogap formed by the assembly of split FP fragment between AuNPs in
nanoclusters was measured from TEM images. A Gatan software (GMS-3) was used to analyze
the intensity profile of the AuNPs in images. The 50
th
percentiles of the average of the maximum
and minimum electron transmission intensities were used to define the edge of each AuNPs at the
gap (Figure 3.14). 314 observations were used to analyze the gap sizes, which followed a
Gaussian distribution with a mean of 2.1 nm and a standard deviation of the mean of 0.5 nm.

Raman spectroscopy
A Horiba, XploRA One microscope with a cuvette holder was used to take liquid Raman
spectra of highly concentrated flGFP (310 μM), flYFP (232 μM), flCFP (363 μM) solutions
(Figure 3.16). Samples were loaded in a quartz cuvette (Starna) and Raman spectra were acquired
for 150 sec using 785 nm laser excitation at 3.33 mW/µm
2
. A TNG buffer blank correction was
applied to each spectrum. The same instrument was used for liquid SERS spectra of sGFP-
AuNPs, M3-AuNPs and AuNP clusters (3.9x10
11
AuNPs/ml) or AgNP clusters (3.1x10
11

AgNP/ml) and for liquid Raman spectra of flGFP and NaPT or TNG buffers (Figure 3.17b and
Figure 3.19), but spectra were taken for 30 sec using 785 nm excitation at 20 mW/µm
2
for AuNPs
and full length GFP or taken for 30 sec using 532 nm excitation at 17 mW/µm
2
for AgNPs.
Liquid Raman spectra of the anionic form of the flGFP chromophore was measured in TNG at pH
8.0. That of its neutral form was measured after dialysis (Slide-A-Lyzer 20K dialysis cassettes,
Thermo Scientific) of the flGFP overnight at 4 ˚C against sodium acetate buffer at pH 6.0. Liquid
SERS spectra of sGFP-AuNPs, M3-AuNPs and AuNP or AgNP clusters were measured in NaPT
buffer pH 8.0 (anionic) or after spinning down the NPs for 8 min at 8000 g and exchanging the
NaPT buffer at pH 8.0 with the same volume of NaPT buffer at pH 6.0 (neutral).
156

For SERS measurements on 5 nm silver island plasmonic substrates, a Renishaw inVia
confocal Raman Microscope was used. SERS spectra of flGFP, sGFP and M3-complemented
sGFP were acquired for 60 sec using a 532 nm laser excitation at 140 µW/µm
2
. SERS spectra of
M3-AgNPs incubated with sGFP were acquired for 30 sec using a 532 nm laser excitation at 140
µW/µm
2
.

Streptavidin titration on biotinylated M3-AuNPs and sGFP-AuNPs
To assess the in vitro binding of biotinylated and non-biotinylated M3-AuNPs and sGFP-
AuNP to avidin, AuNPs were incubated with decreasing concentrations of streptavidin (Sigma,
final concentrations: 3.33 μM, 0.83 μM, 0.42 μM, 0.17 μM, 41.67 nM, 25 nM, 16.67 nM, 3.33
nM) for 45 minutes at room temperature. Samples were run in 0.8 % agarose gel at 50 V in TAE
buffer pH 8.0 (Figure 3.23). The first lanes represent control samples without streptavidin.
Shifted bands indicate that streptavidin only reacts with biotinylated M3-AuNPs or sGFP-AuNPs.

Fluorescence, dark field and total internal refection fluorescence microscopy imaging
of AuNPs targeted to avidin biomarkers in cells
HeLa or U2OS cells were grown at 37 ˚C on borosilicate coverslips (Marienfeld, 25 mm
diameter, #1.5 thickness) in DMEM media (Lonza) supplemented with 10 % fetal calf serum
(FCS). Cells were transiently transfected with cDNA coding for the transmembrane or the GPI
avidin fusions for 24 hours (XtremeGene, Roche). 4 hours prior to incubation with biotinylated
M3-AuNPs or sGFP-AuNPs, cells were starved in FCS-free DMEM at 37 ˚C to free the avidin
fusions from biotin present in the FCS supplement and to avoid competition with the biotinylated
NPs. Starved cells were incubated with sGFP-AuNPs or M3-AuNPs separately for 1 hour at 37
157

˚C at a concentration of 0.57x10
10
AuNPs/ml (dark field imaging) or 1.43x10
11
AuNPs/ml (total
internal refection imaging).
For correlated dark field imaging of targeted AuNPs and fluorescence imaging of residual
avidin fusions at the cell plasma membrane, cells were additionally incubated with 2 µM of
biotin-Alexa594 for 10 min at 37 ˚C before multiple rinses in PBS and cell fixation in 4 %
paraformaldehyde for 15 min. Microscopy imaging was done in PBS on an inverted Nikon
Eclipse Ti-E microscope equipped with a Plan Fluor ELWD x40 objective (Nikon), a condenser
lens, a mercury lamp, appropriate optical filters for Alexa594 imaging (Exc:562DF40, Dichroic:
593-Di03 and Em: 641DF75, Semrock) and an Ixon Ultra EMCCD camera (Andor).
Total internal refection fluorescence (TIRF) microscopy of M3-AuNPs or sGFP-AuNPs
targeted at the bottom membrane of expressing cells was performed on the same inverted Nikon
Eclipse Ti-E microscope equipped with a x100, 1.49 NA objective, TIRF optics, a 561 nm laser
line, appropriate optical filters (Exc: ZET405/488/561/647x, Dichroic: ZT405/488/561/647 and
Em: 600DF50, Chroma) and an Ixon Ultra EMCCD camera (Andor). Images were acquired at
100 ms/frame.

Scanning Electron Microscopy
Cells were grown at 37 ˚C on coverslips (Neuvitro, 15 mm diameter, #1 thickness) and
transiently transfected with cDNA coding for the transmembrane and the GPI avidin fusions as
described above. After 4 hours starving in FCS-free DMEM, cells were incubated with 1.07x10
10

NPs/ml of biotinylated M3-AuNPs and 1.07x10
10
NPs/ml of biotinylated sGFP-AuNPs separately
or simultaneously for 1 hour at 37 ˚C. After a washing step with NaP buffer, ½ strength
Karnovsky’s fixative was immediately applied for 1 hour at room temperature to fix the cells. 0.1
M Cacodylate buffer was used to rinse the Karnovsky’s fixative and cells were treated with 2 %
158

osmium tetroxide for 30 min at room temperature for heavy metal staining. Increasing
concentrations of ethanol were then applied for a gradual dehydration of the specimens, which
were allowed to air-dry overnight at room temperature after application of hexamethyldisilizane.
Images were acquired on a JEOL JSM-6390LV scanning electron microscope at 10 kV, with 5
mm working distance and a 30 nm spot size.

SERS microscopy of AuNPs in cells
For fixed cells, cells were grown at 100 % confluence on coverslips (Neuvitro, 15 mm
diameter, #1 thickness), transfected, starved and incubated with 1.07x10
10
NPs/ml of biotinylated
sGFP-AuNPs and 1.07x10
10
NPs/ml of biotinylated M3-AuNPs separately or simultaneously as
described above. Fixation was performed for 15 min at room temperature with 2 %
paraformaldehyde in NaP buffer, after washing the cells. The coverslips were then mounted on
SiO 2 wafers (300 nm oxide thickness, University Wafer) in NaP buffer pH 8.0 with 5 % glycerol.
A wide-field Rima Hyperspectral Imaging System (Photon Etc) equipped with a x100 objective,
Bragg tunable filters and an EMCCD was used to map 130x130 μm
2
areas of the cell samples
with a 532 nm laser excitation at 200 µW/µm
2
. Each image was acquired for 60 seconds with a 3
cm
-1
spectral resolution. SERS spectra and SERS images of cells at specific Raman shifts were
reconstructed with a PHysSpecV2 software (Photon Etc).
For live cells imaging, cells were grown at 100 % confluence directly on SiO 2 wafer (300
nm oxide thickness, University Wafer), transfected, starved and incubated with biotinylated
sGFP-AuNPs (1.07x10
10
NPs/ml) and biotinylated M3-AuNPs (1.07x10
10
NPs/ml) or with
biotinylated sGFP-AgNPs (8.56x10
9
NPs/ml) and biotinylated M3-AgNPs (8.56x10
9
NPs/ml)
simultaneously as described above. After 1 hour incubation with NPs, cells were washed with
PBS (8 mM NaH 2PO 4, 150 mM NaCl) at pH 8.0 and imaged in this buffer over a one-hour
159

period. To stimulate early endocytosis, the PBS buffer was replaced by a hypotonic PBS buffer at
pH 8.0 (8 mM NaH 2PO 4, 50 mM NaCl). Confocal imaging was performed on a Renishaw inVia
Raman Microscope equipped with a x40 water immersion objective. Cell samples were raster
scanned in 1 μm step sizes using a 532 nm laser excitation at 6.73 mW/µm
2
and using a
cylindrical lens to spread the laser spot into a 70 μm × 1 µm line. At each scanning steps, signal
integration was performed for 30 sec with 1.3 cm
-1
spectral resolution. SERS spectra and SERS
images of cells at specific Raman shifts were reconstructed using a home-made Matlab program.

Photoacoustic microscopy imaging of AuNPs clusters on targeted cells
Cells were grown at 100 % confluence on coverslips (Marienfeld, 25 mm diameter, #1.5
thickness), transfected, starved and incubated with 3x10
10
NPs/ml of biotinylated sGFP-AuNPs
and 3x10
10
NPs/ml of biotinylated M3-AuNPs separately or simultaneously as described above.
After a washing step with NaP buffer, cells were immediately fixed in ½ strength Karnovsky’s
fixative for 1 hour at room temperature. Fixed cells were immersed in PBS buffer for acoustic
signal coupling. Imaging was performed on a custom photoacoustic microscope as previously
described [148, 275]. Briefly, the laser source is a diode-pumped solid-state Nd:YAG laser (Spot-
10-200-532, Elforlight Ltd) with a wavelength of 532 nm and a pulse duration of 2 ns. The laser
was first collimated by a lens system, then reflected by a 2D galvanometer (6230H, Cambridge
Technology), and, finally, focused on the sample by an achromatic objective (AC254-040-A,
Thorlabs) with a focal length of 40 mm and a numerical aperture (NA) of 0.1. Excited acoustic
signals were captured by a custom ultrasonic transducer (center frequency: 35 MHz, 50 %
bandwidth at -6 dB), amplified by a low-noise amplifier (ZFL-500LN, Mini-Circuit), digitized by
an A/D card (Cobra CompuScope CS22G8, GaGe), transferred to the computer, and, finally,
reconstructed using a maximum amplitude projection (MAP) algorithm for visualization. For
160

photoacoustic imaging and signal intensity measurements at fixed laser excitation (Figure 3.30a,
b), a 532 nm laser pulse excitation energy of 130 nJ was used. For signal quantification, the
means and standard deviations of the total photoacoustic amplitudes were calculated over a
similar number of scanning areas totaling 3 mm
2
of cells at 100 % confluence. The mean total
photoacoustic signal of transfected cells not incubated with AuNPs was used for background
correction.

161

CHAPTER 4. FUTURE DIRECTIONS
Introduction
Since its discovery [5, 6], the use of Raman spectroscopy has been growing
exponentially. Ongoing studies provide deeper insight and open up opportunities for new
strategies to improve the SERS nanoprobes, the detection sensitivity and selectivity, or the
Raman instrumentation and multimodality that are appropriate for in vivo applications. In this
chapter, the potential future steps related to this thesis work are described.

SERS nanoprobes
Nanomaterial shape is one of the important parameters to optimize the electromagnetic
(EM) field enhancement [22]. The enhancement factor (EF) of isolated single gold nanospheres
has been previously reported as 10
3
-10
4
[276]. On the other hand, single gold nanostar has
exhibited 10
7
enhancement of Raman scattering [277]. In this thesis, we showed ~10
6
EF from
assembled dimeric gold nanoparticles which is consistent with our FDTD simulation study [253].
The sharp edges and tips of gold nanostars provide highly localized electromagnetic field where
the highest EF is obtained [34]. It has been shown that the spike number of the nanostar can be
easily tuned and it closely relates to the LSPR frequency of the gold nanostar [278]. Simply, as
the number of spike number increases, the more red-shifted LSPR frequency is observed.
Because the number of spike and their morphology contribute to the plasmon band position, large
LSPR frequencies (~700-800 nm) can be obtained from gold nanostars as small as ~30 nm
diameter. These features make them promising SERS probes for cellular and in vivo applications.
Complementary sFP and M3 peptide fragments can be used to modify the gold nanostars and due
to the higher spike number, the probability to position the complementary fragments on the tips of
162

the spikes gets higher. When guided gold nanostar clusters are formed, nanostar assemblies might
produce higher electromagnetic field enhancement compared to gold nanoparticles.

Figure 4.1 Colloidal gold nanostars. (a) Vis/NIR spectra of gold nanostars synthesized by gold nanoparticle (30nm)
seeding. R is the ratio of gold salt concentration to gold nanoparticle seed concentration. (b-d) TEM images of gold
nanostars at R= 1.67, 11.25, and 20, respectively [278].

Detection sensitivity and selectivity
In this thesis work, targeting of live cells were achieved by transient transfection of
transmembrane and GFP anchored avidin fusions. Basically, we increased the number of
targeting receptor on the cell plasma membrane in order to increase the probability of SERS
probe binding events. In in vivo applications, this situation can be improved by dual targeting of
two natural receptors on the cell plasma membrane. Dual targeting might provide both higher
efficiency of binding and higher detection sensitivity and selectivity. Under dual targeting
conditions, each complementary fragment modified gold nanomaterials react with different
surface receptors, therefore this might enable more interaction between the probes and the cell
surface due to the increased receptor population. More importantly, the dual interaction gives rise
163

to a higher detection sensitivity due to the required complementation of sFP-M3 peptide
fragments to observe the Raman fingerprints. In the absence of any targeting receptors on the cell
plasma membrane, Raman signal of the reporter cannot be detected, therefore the detection
system becomes more sensitive for targeted cells. Furthermore, the necessity of concomitant
presence of two receptors increases the selective recognition of the targeted cells among other cell
types. Here we showed a potential use our novel SERS nanoprobes to detect targeted cells via
noninvasive Raman imaging method. As a future step of this thesis work, these nanoprobes can
be applied to various cancer cell models with different receptors. This can further prove the
capability of this system for the detection of different cancer types.

Raman instrumentation and multimodality
As mentioned in CHAPTER 1 (1.3.3), hyperspectral Raman imaging with angle-tunable
filters provides faster image acquisition at Raman band of interest compared to raster scan where
the entire spectra have to be collected to reconstruct the image at particular Raman shift [150].
Also, 2D-galvano mirrors have been implemented to the light path for rapid laser scan on a fixed
stage which facilitates in vivo animal imaging. The former method provides a faster spectral
scanning, the latter method provides a faster laser scanning. These two methods can be combined
and outcome system might provide even faster in vivo animal or cellular Raman imaging systems.
The combined system might be beneficial particularly for live in vivo imaging where
simultaneous feedback from the tissue holds great importance such as surgical applications.
Furthermore, fiber-optic based in vivo Raman spectrometer systems hold great potential
as a diagnostic tool [279]. In general, fiber-optic probes consists of two fibers, delivery fiber
terminates at laser and collection fiber terminates at spectrometer (Figure 4.2). These two fibers
meet at the bi-furcation point which terminates at the gauge needle tip. Instead of spectrometer at
164

the collection fiber terminus, angle-tunable filter set and a CCD camera can be implemented. The
combination of fiber-optic probe and hyperspectral Raman imaging might provide faster
acquisition from in vivo endoscopic applications.

Figure 4.2 Illustration of general architecture of fiber-optics probes with bifurcation point [279].

Stimulated Raman scattering (SRS) microscopy employs simultaneously two laser beams
on a sample [280]. This system has been developed as a label-free biomedical imaging technique.
When the frequency difference of the two laser beams (“pump” and “stokes” lasers at ω p and ω s,
respectively) matches the Raman band of interest, the amplification of the Raman band occurs.
However, if the frequency difference does not match any vibrational modes, the stimulated
excitation does not occur, therefore non-resonant background effect is eliminated. The SRS
system might be performed in the presence of our SERS probes where the pump laser can be used
at the LSPR frequency of gold nanoparticles and the stokes laser can be used at one of the
complemented split-GFP Raman modes (e.g. 1527 cm
-1
). As a result, a higher signal to noise
detection might be achieved while avoiding the strong physiological background.
Multimodality is a critical characteristic of nanoprobes since in general, multiple
techniques are necessary for sensitive diagnosis and treatment. Nanomaterials can be
functionalized as a common contrast agent for multiple diagnosis and treatment techniques. These
nanomaterials are called “theranostics”. As mentioned earlier, photoacoustic imaging and
165

photothermal therapy are complementary techniques where a single nanoprobe can provide
improved photoacoustic signal intensity and enhanced thermal expansion for photothermal
therapy. Additionally, SERS imaging can be implemented into these two complementary
techniques. In this tri-modality imaging system, our SERS nanoprobes offers (i) single cell
detection sensitivity via SERS imaging; (ii) higher photoacoustic signal due to the formation of
clusters on the cell plasma membrane; (iii) enhanced thermal expansion due to the metallic
nanoparticles. Acquiring the SERS signal before and after the photothermal treatment can provide
further information about the tumor area. As a next step, these three systems can be combined and
our SERS nanoprobes might be good candidates for in vivo animal applications.

Summary and Conclusion
Guided clustering of metal nanoparticles using protein scaffolds is a promising strategy to
obtain high signal to noise SERS detection. As discussed in CHAPTER 1, Raman spectroscopy
provides chemical information of molecules, although its cross section (~10
-30
) is significantly
lower than fluorescence cross section (~10
-16
). Therefore, observing a Raman scattered photon is
highly challenging in the presence of background noise. This method can be massively enhanced
by plasmonic metal nanomaterials, which are easily utilized for biomedical applications. Surface
modification of metal nanoparticles becomes critical to keep nanoparticles stable in physiological
environments. Covalent-like thiol-metal interactions have been widely used to form a stable
surface modification. PEG chains are also another important component of the metal nanoparticle
surfaces because these versatile ligands make nanoparticles more biocompatible and more
soluble. Furthermore, there is a variety of Raman reporters and it is essential to select reporters
that are well suited for a given application. In this thesis work, we focused on split green
fluorescence protein fragments as a Raman reporter upon complementation. In CHAPTER 2, the
166

characterization of split-FP fragments and their self-assembly process were described
comprehensively. We made point mutations on sGFPori structure to improve its folding and
chromophore maturation rates as well as to create spectral variants of sGFPori. Furthermore,
these split-FP variants exist as monomers at the cell plasma membrane and suitable for single
molecule and confocal imaging. In in vitro experiments, sGFP2 and flGFP2 variants exhibited
higher quantum yield and brightness compared to the sGFPori. Also, this sGFP2 variant has an
improved maturation rate compared to sGFPori. As discussed in CHAPTER 3, we used sGFP2
and the complementary M3 peptide fragment to modify metal nanoparticles. Autocatalytic
cyclization of these fragments formed dimeric metal assemblies and the complemented sGFP2, as
a Raman reporter, is positioned at the nanogap. The created plasmonic hot spots confined the
electromagnetic field, which enhanced the Raman fingerprints of the complemented sGFP2. In
CHAPTER 1, we explained that there are many areas where metal nanoparticles can be used,
such as biosensing, biomedical imaging, or theranostic and therapeutic applications. Here, we
used our SERS probes to detect targeted cells by Raman imaging and photoacoustic microscopy.
Besides the split-FP fragments, biotinylated PEG chains were used on the metal nanoparticle
surface to specifically detect transmembrane and GPI-anchored avidin fusion expressed in live
cells. We showed that our SERS probes selectively detect avidin expressing cells and upon in situ
assembly of the nanoparticles we obtained SERS signals from both live and fixed cells.
Photoacoustic microscopy provided further evidence of the specific assembly with increased
signal intensity due to the formation of nanoparticle clusters at the cell membrane.
To conclude, in this thesis, we present a novel technique to assemble metal nanoparticles
into photonically active plasmonic hot spots by split fluorescent protein scaffolds. We show the in
situ assembly of metal nanoparticles enables the detection of the targeted live cells by Raman
imaging. We believe the findings presented in this work offer important insights into the
167

photophysical characteristics of split-FPs, the self-assembly process of split-FPs, and the use of
split-FPs both as a Raman reporter and as a scaffold agent to create plasmonic nanoassemblies for
biomedical applications.

168

Instrumentation for Raman Spectroscopy
A1.1.1 Continuous Lasers
Raman spectroscopy has become a general analytical technique after the advent of the
laser in 1960. There are important parameters of laser source for Raman systems, such as; laser
wavelength, linewidth, output power, polarization, and beam quality. In principle, any laser
excitation wavelength can be used in Raman systems because the Raman spectroscopy measures
the frequency shift from the excitation wavelength. However, shorter wavelengths of the visible
range are the main preferences because the Raman scattering efficiency and the fourth power of
the excitation frequency (ω
4
) are directly proportional. Using short wavelengths in Raman
systems can cause fluorescence due to high photon energy which can dominate the Raman signals
and reduce high signal to background detections. To overcome this fluorescence background,
NIR wavelengths can be used since they do not have enough energy to populate the excited
electronic states and induce fluorescence. In biological applications, the NIR wavelengths hold
further importance to obtain the minimal absorption by hemoglobin and water, and this allows the
highest tissue penetration [161]. The main problem of using NIR excitation wavelength is that the
lower the frequency (ω
4
), the harder the detection of Raman signals becomes because these
Raman bands will be shifted to infrared (IR) range where classical detector have poor detection
quantum efficiencies. However, improved CCD detectors, holographic notch filters, and
holographic gratings make the detection of Raman signals with NIR excitation possible. The
required output power to obtain Raman signals is dependent on the sample type. In general, the
Raman signal intensity is proportional to the excitation laser power. High laser power shortens
the acquisition time, and increase the signal to noise (S/N) ratio. As long as the sample is not
damaged, laser power range can be a few microwatts to tens of watts. For sensitive samples,
169

microwatt powers can be used with longer integration times (1-10 seconds) to obtain high S/N
ratios. The resolution of Raman spectrum can be affected by laser linewidth, grating resolution,
and the linewidth of the probed vibrational transitions. Most commercial continuous lasers
provide narrower linewidth characteristic than the grating resolution and the linewidth of the
probed vibrational transition. Typically, a laser with 1 cm
-1
linewidth is sufficient to couple with a
CCD detector at 2-6 cm
-1
spectral resolution. The spot diameter of the laser beam describes the
beam quality. A good quality laser beam usually has a nearly diffraction limited, 1-2 μm spot
diameter. Laser polarization becomes important while working with crystalline materials to
determine the orientation of the crystals. Moreover, the polarization of the output laser beam can
be used to obtain confined electromagnetic field enhancement from nanostructures on thin films.

A1.1.2 Holographic notch filters
Holographic optics are critical components of Raman systems. The application of
holographic notch filter for rejection of Rayleigh scattering was shown as an inexpensive and
efficient way for band-rejection in Raman systems [281, 282]. The improved holographic notch
filters have sharper spectral edges and narrower spectral bandwidth compared to classical filters.
These characteristics provide the rejection of wavelengths in a narrow range, and the collection of
Stokes and anti-Stokes Raman bands close to the Rayleigh line [283]. Therefore, working with
low power laser sources becomes possible where the combination of highly reflected laser and
highly transmitted Stokes shifts is utilized. Furthermore, holographic beam-splitters also increase
the efficiency of Raman systems. They are used as 90˚ reflector which directs the laser output
onto the sample and transmits the Raman scattering while rejecting the laser backscattering.

170

A1.1.3 Holographic gratings
A grating splits and diffracts the light into different wavelengths. Holographic gratings
provide a sinusoidal groove pattern and high groove density which increase the light collection
efficiency. The efficiency of light collection is defined by the f-number of the grating which is
also related with aberration correction. Compared to the standard ruled grating, aberration
corrected holographic gratings have higher f-number with high groove densities that can cover
UV/NIR region.

A1.1.4 Array detectors
High throughput and sensitive detectors have been critical components to convert Raman
techniques into common analytical tools. Raman spectroscopy requires detector systems with low
noise characteristics and high quantum efficiency. Array detectors are made of an array of pixels
where each pixel defines the resolution of the detector. They can record all wavelength
simultaneously which decreases the integration time of data acquisition. Charged-coupled device
(CCD) has two-dimensional array of pixels; one is for frequency detection and the other is for
spatial resolution which is required for map scan Raman analysis.

Figure A1.1 Typical setup for Raman spectroscopy equipped with holographic components, CCD detector, continuous
wavelength laser.

171

A1.1.5 Renishaw inVia Raman system
This Raman microscope is the one most commonly used in this thesis. The system is
equipped with a 532 continuous laser, a holographic notch filter, a holographic grating, and a
CCD camera. Mainly, two different objectives are used; 40X, NA 0.6 air objective and 40X, NA
1.15 water immersion objective. A cylindrical lens is included into the light path, before the notch
filter to expand the beam for line scans. Based on the used objective, the applied laser output
power intensity on the sample varies between 1- 50 μW/μm
2
. The laser spot diameter is 1.08 μm.
The holographic grating coupled with 532 nm laser wavelength has 1800 l/mm groove density.

Figure A1.2 Renishaw inVia Raman microscope system.

172

Surface Chemistry of Metal Nanostructures
A2.1.1 Surface modifications
DNA wraps around the nanoparticle surface at low surface coverage, and short strand
DNA aligns perpendicularly to the surface and stretches out, whereas long strand DNA exhibits
the same behavior but at the outer region it creates random coil-shaped configuration at the high
surface coverage. Higher surface coverage forms tighter ligand coating which supports better
stability at high ionic strength conditions [84].

A2.1.2 Poly(ethylene) Glycol (PEG) coating
In order to improve the long-term stability of these nanoparticles, block copolymers
consist of multiple sulfide groups, poly(propylene sulphide) (PPS), with PEG were developed
[66]. In this method, sulfide groups on the block copolymer chemically adsorb on gold
nanoparticles, and PEG groups provide passivation of the surface against nonspecific protein
adsorption. The organic adlayer, PEG-bl¬-PPS-bl-PEG, have performed greater oxidative
stability due to the multi-sulfide bearing PPS polymer, and improved surface passivation due to
the PEG polymers. On the other hand, to avoid the thiolate oxidation, it has been shown that
block copolymers without any sulfide groups, poly((2-N,N-dimethylamino)ethyl methacrylate)
(PAMA), can be conjugated to PEG groups where oligoamine segments were employed to form
N-Au linkage [284]. Even though oligoamine segments provide multipoint coordination on gold
nanoparticle surface, the improved stability of this system (PEG-bl-PAMA) depends on the
increased surface density of PEG polymers with shorth length PAMA chain.

173

A2.1.3 Hot spot formation of metal nanoparticles
Another strategy is block polymer encapsulation to protect NaCl driven multimeric
nanostructures and correlate these nanostructures with the correct SERS enhancement. In this
method, formed nanoaggregates (dimers, trimers, tetramers) after the Raman reporter adsorption
(i.e. 2-naphthalenethiol) are embedded by polystyrene-block-poly(acrylic acid) (PSPAA) (Figure
A2.1) [285].

Figure A2.1 PSPAA encapsulation on gold nanostructures functionalized with 2-naphtalenethiol. TEM images of (a)
monomeric, (b) dimeric, and (c) trimeric gold nanostructures with PSPAA layer (gray ring). (d) UV/Vis spectra of
PSPAA encapsulated gold nanostructures, and differential centrifugation result to isolate multimeric nanostructures
groups [285].

Bonham et al. has applied DNA scaffolding technique to detect proteins [286]. In this
study, specific DNA sequence has been encoded in linking DNA duplex that can recognize
certain proteins which are labeled with Raman reporters. DNA hybridization brings gold
nanoparticles in close proximity and creates hot spots which gives rise to SERS enhancement.
DNA hybridization driven hot spot formation has become a highly tunable system with more
174

complex surface functionalization methods. Lim et al. has shown programmable assembly by
Cy3-conjugated DNA hybridization on gold nanoparticles [242]. After hybridization, silver shell
has been grown on dimeric structures. Formation of thin silver shell around the hot spot enhances
the SERS signal of the Raman reporter, Cy3. These dimeric Au/Ag core-shell nanospheres have
been highly reproducible and have a potential use in bioapplications. Although this system is
quite efficient, multiple point gap junctions cause non-uniform SERS signals.
DNA scaffolding methods facilitate controllable building of 2D, 3D, and more complex
structures, however these systems require conjugation of external Raman reporters. A promising
approach to overcome this issue is to employ Raman reporters as a scaffold to bring nanoparticles
in close proximity and create well-defined nanogaps. Taylor et al. has reported the use of
cucurbit[n]uril (CB[n]) macrocyclic host molecule as a molecular glue that can produce fixed and
rigid, 0.9 nm spaced hot spots between gold nanoparticles [238]. CB[5] and CB[7] molecules
have rigid barrel shape structures, they are water-soluble, and Raman-active. These molecules
bind to gold nanoparticles through carbonyl groups and can hold a guest molecule, via a
molecular recognition based interaction, in the barrel shape geometry. Therefore, the host-guest
complexation within the hot spot advances the solution based, multifunctional SERS sensing
applications. Although this approach has multiple advantages in terms of the rigid hot spot
formation and the use of dual functioning Raman reporter, the self-assembly of nanoparticles in
cells is the major drawback of this method.

175

SERS nanoprobes for biosensing
Different surface modification methods facilitate SERS based ultrasensitive single
molecule detection sensitivity through surface appended Raman reporters. It has been shown that
the deformation of the benzene ring in the Raman reporter causes a Raman band shift which can
be utilize for the detection of H1 influenza protein at picomolar sensitivity, a significant
improvement over the nanomolar sensitivities obtained from ELISA tests [287]. In this method, a
dielectric glass surface is covered with Ag film which is functionalized with 4-aminothiophenol
(4-ATP) conjugated to an antibody to recognize the H1 influenza protein. The antibody-antigen
interaction creates a mechanical deformation on the benzene ring, hence the Raman band shifts.
Recently, a similar method has been applied for multiplex detection of different serological
cancer markers on the same quartz slide chip [288]. In this system, each row of wells in the quartz
chip was functionalized with three different antibodies. Gold nanostars were functionalized with
4-nitrobenzenethiol (4-NTP) Raman reporter, then encapsulated with silica shells. The amine
chemistry was used to modify the silica shells to create three different gold nanostars with
different antibodies. After the application of serum on the chip, gold nanostars were introduced
and the antibody-antigen-antibody sandwich model created the Raman fingerprints of 4-NTP
reporter.

SERS nanoprobes for biomedical imaging
Multiple cytotoxic effects of carbon nanotubes have been reported, since non-
functionalized single- or multi-walled carbon nanotubes (SWNTs or MWNTs) are insoluble. On
the contrary, functionalized SWNTs and MWNTs becomes stable in aqueous environments,
hence they do not exhibit any toxicity in cell cultures [289]. From this standpoint, it is hard to
176

compare these studies, because they are not conducted under the same conditions, such as the size
of the nanotubes, the functionalization, the cell culture type, the organs that are treated with
carbon nanotubes are quite different in each study.
Furthermore, different geometries of gold nanostructures have been also used for in vivo
PA imaging studies [290]. In this study, platinum (Pt) cubic core with gold spheres has been
synthesized to form hetero-tripod nanostructures which provide both visible and NIR LSPR
frequencies. Au-tripods were functionalized with PEG-conjugated RGD molecules to detect
glioblastoma tumor and a radioactive metal chelator, NOTA, to track the RGD-Au-tripods in vivo
for pharmacokinetic and biodistribution analysis by PET. Even though the tumor uptake increases
over time, RES system organs, liver-kidney-spleen, show accumulation suggesting the immune
system excretion of the Au-tripods. Also, increased kidney retention shows these particles can
also be excreted through renal system. A linear correlation between tumor uptake of the RGD-
Au-tripod and the photoacoustic signal intensity has been observed. Most interestingly, when the
targeted integrins on tumor cell membrane have been blocked, the low PA signal from the tumor
exhibits the significantly decreased cellular uptake of RGD-Au-tripods which suggests the
targeted Au-tripods can detect the tumor tissue selectively (Figure A3.1a). Furthermore, PA
imaging results have shown a good correlation with PET images indicating PA imaging can be
used to identify functional and molecular details about the tumor tissue with high spatial
resolution (Figure A3.1b). Although the radioactive metal chelator has been used in this study to
track the Au-tripods, Raman reporters can be used instead of radioactive labels to chase the
plasmonic nanostructures via SERS imaging systems.
177

Figure A3.1 Photoacousting imaging and PET scans on tumor that are treated with Au-tripods. (a) Photoacoustic
images (PAI) of targeted integrins on tumor cell membrane pre- and post-injections of RGD-Au-tripods at 1h, 2h, 4h
time points (left panel). After blocking the integrins on tumor cell membrane, photoacoustic images of pre- and post-
injections of RGD-Au-tripods. (b) Small animal PET images of integrin positive tumor bearing mice injected with
RGD-Au-tripods to target integrins [290].

SERS nanoprobes for therapeutic and theranostic applications
Instead of passive accumulation based on leaky tumor vasculature, immunotargeted
nanoshells have been engineered to absorb and scatter at NIR range (~800 nm) [291]. Human
breast adenocarcinoma cells that overexpress HER-2 have been treated with anti-HER-2
functionalized nanoshells and imaged using darkfield imaging microscopy. Upon NIR irradiation,
cells have been treated with calcein dye to identify cell viability, and with silver staining to show
the presence of nanoshells (Figure A3.2). The results have proven the multimodality of targeted-
nanoprobes in cancer therapy and bioimaging applications.
178

Figure A3.2 Nanoshells/NIR laser combination with dark field imaging on breast adenocarcinoma cells expressing
HER-2. Scattered-based dark field imaging to assess HER2 expression (top row), calcein staining to analyze cell
viability after laser treatment (center row), silver staining to localize nanoshells bindings (bottom row). In dark field
imaging, anti-HER2 modified nanoshells is observed, this indicates cells with HER2 expression can be targeted with
nanoshells (top row, right column). After NIR laser treatment, cytotoxicity (dark spot) is observed only from nanoshells
treated sample (center row, right column). Silver staining reveals the presence nanoshells on the cells (bottom row right
column) [291].

Later, nanomatryoshka probes have been designed to possess superior theranostic
characteristics than the nanoshells (Figure A3.3) [121]. Nanomatryoshkas are multilayered gold
nanoparticles with a gold core, a silica shell, and a top gold layer. Therefore, these nanoprobes
have two distinct LSPR frequency, 560 nm and 783 nm. Comparative study of photothermal
transduction have shown that nanomatryoshkas make a better conversion from light to heat than
nanoshells. Triple negative breast cancer (TNBC) tumor-bearing mice have been used as a model
system. As a result of biodistribution analysis, nanomatryoshkas have penetrated to the tumor
tissue five times more than nanoshells due to the size effect. This cancer cells have been
transfected with luciferase bioluminescence agent to visualize the in vivo tumor tissue after the
nanoprobe/NIR irradiation. Nanomatryoshka/NIR treatment has entirely ablated the tumor cells,
whereas nanoshells/NIR treatment has shown recurrence of tumor overtime due to the lack of full
179

eradication of cancer cells (Figure A3.3). Multilayered nanoprobes have also potential to use as
NIR photothermal agents against aggressive tumors.

Figure A3.3 Tumor tissue response to photothermal therapy monitored by bioluminescence imaging. Luciferase
transfected tumor cells create bioluminescence signal. Representative mice of each condition for certain days of the
evaluation. Nanomatryoshka (NM) injected mice do not show tumor tissue recurrence even 60 days after the laser
treatment. However, nanoshells (NS) injected mice show tumor recurrence 2 weeks after the laser treatment. [121].

Recently, targeted gold nanobipyramids (AuNBPs) conjugated with Raman active 2-
naphthalenethiol (2-NT) has been used for SERS detection of MCF-7 breast cancer and
photothermal ablation of the tumor [292]. AuNBPs have been first functionalized with 2-NT, then
SH-PEG-NH 2 heterofunctional linker in where amino group has been covalently reacted with
carboxyl group of folic acids (FA) that specifically recognize MCF-7 breast cancer cells. In vitro
experiments have shown the cell viability/cell death by calcein AM/ethidium homodimer-1
staining and a clear photothermal ablation result has been obtained upon NIR (808 nm) irradiation
of FA-NT-AuNBPs treated MCF-7 cancer cells (Figure A3.4a). After the injection of targeted
180

nanoprobes, tumor-bearing mice have been scanned and Raman fingerprints of 2-NT have been
observed in the tumor site. In vivo IR thermal images of MCF-7 tumors have been taken to
evaluate the temperature elevation under NIR irradiation (Figure A3.4b). In 5 min, temperature
increase at the tumor site has reached to above 70 ˚C which is sufficient to ablate the tumors.
Additionally, the digital pictures of the nude mice have shown the photothermally treated visible
tumor-bearing mice at the first day and the fourteenth day of the treatment (Figure A3.4c). The
results clearly indicate the FA-NT-AuNBPs can eradicate the tumor cells under NIR irradiation
and provide a platform for SERS imaging.

Figure A3.4 In vitro and in vivo characterization of FA-NT-AuNBPs treated MCF-7 cells and MCF-7 tumor bearing
mice. (a) Fluorescence images of MCF-7 cells (A) without and (B) with FA-NT-AuBPs treatment under 808 nm laser
irradiation at different time points. Calcein AM/ethidium homodimer-1 staining is used to differentiate live/dead cells.
181

Calcein AM with green fluorescence indicates live, ethidium homodimer-1 with red fluorescence indicates dead cells.
There is a clear photothermal ablation with FA-NT-AuBPs treatment upon 808 nm laser irradiation. (b) IR thermal
imaging of MCF-7 tumor bearing mice with (A) PBS control injection and (B) FA-NT-AuBPs injection under 808 nm
laser irradiation at 0 min, 1 min, 3 min, 5 min. Fa-NT-AuBPs injected mice group shows a significant temperature
increase in the tumor region upon laser irradiation. (c) Digital photographs of MCF-7 tumor bearing mice at 0 day and
14 days time points from four different groups; a-PBS, b-PBS+NIR, c-FA-NT-AuBPs, d-FA-NT-AuBPs+NIR [292].

182

BIBLIOGRAPHY
1. Raman C. V., K.K.S., A New Type of Secondary Radiation Nature, 1928. 121: p. 501-
502.
2. Wang, Y. and R.L. McCreery, Evaluation of a diode laser/charge coupled device
spectrometer for near-infrared Raman spectroscopy. Analytical Chemistry, 1989. 61(23):
p. 2647-2651.
3. Williamson, J.M., R.J. Bowling, and R.L. McCreery, Near-Infrared Raman Spectroscopy
with a 783-nm Diode Laser and CCD Array Detector. Applied Spectroscopy, 1989.
43(3): p. 372-375.
4. Nie, S. and S.R. Emory, Probing Single Molecules and Single Nanoparticles by Surface-
Enhanced Raman Scattering. Science, 1997. 275(5303): p. 1102-1106.
5. Fleischmann, M., P.J. Hendra, and A.J. McQuillan, Raman spectra of pyridine adsorbed
at a silver electrode. Chemical Physics Letters, 1974. 26(2): p. 163-166.
6. Jeanmaire, D.L. and R.P. Van Duyne, Surface raman spectroelectrochemistry: Part I.
Heterocyclic, aromatic, and aliphatic amines adsorbed on the anodized silver electrode.
Journal of Electroanalytical Chemistry and Interfacial Electrochemistry, 1977. 84(1): p.
1-20.
7. Philpott, M.R., Effect of surface plasmons on transitions in molecules. The Journal of
Chemical Physics, 1975. 62(5): p. 1812-1817.
8. Smith, W.E., Practical understanding and use of surface enhanced Raman
scattering/surface enhanced resonance Raman scattering in chemical and biological
analysis. Chemical Society Reviews, 2008. 37(5): p. 955-964.
9. Lyon, S.A. and J.M. Worlock, Role of Electromagnetic Resonances in the Surface-
Enhanced Raman Effect. Physical Review Letters, 1983. 51(7): p. 593-596.
10. Creighton, J.A., C.G. Blatchford, and M.G. Albrecht, Plasma resonance enhancement of
Raman scattering by pyridine adsorbed on silver or gold sol particles of size comparable
to the excitation wavelength. Journal of the Chemical Society, Faraday Transactions 2:
Molecular and Chemical Physics, 1979. 75(0): p. 790-798.
11. Raman C. V., K.K.S., Polarisation of Scattered Light-quanta. Nature, 1928. 122: p. 169-
169.
12. Fang, C. and G. Wu, Temporal electronic structures of nonresonant Raman excited
virtual states: a case study of ethylene thiourea. Journal of Raman Spectroscopy, 2007.
38(11): p. 1416-1420.
13. Souza Filho, A.G., et al., Stokes and anti-Stokes Raman spectra of small-diameter
isolated carbon nanotubes. Physical Review B, 2004. 69(11): p. 115428.
14. Balkanski, M., R.F. Wallis, and E. Haro, Anharmonic effects in light scattering due to
optical phonons in silicon. Physical Review B, 1983. 28(4): p. 1928-1934.
15. Efremov, E.V., F. Ariese, and C. Gooijer, Achievements in resonance Raman
spectroscopy: Review of a technique with a distinct analytical chemistry potential.
Analytica Chimica Acta, 2008. 606(2): p. 119-134.
16. Chalmers, J.M., Peter R. Griffiths Handbook of Vibration Spectroscopy. 2002: John
Wiley and Sons.
17. Greenler, R.G. and T.L. Slager, Method for obtaining the Raman spectrum of a thin film
on a metal surface. Spectrochimica Acta Part A: Molecular Spectroscopy, 1973. 29(1): p.
193-201.
183

18. Albrecht, M.G. and J.A. Creighton, Anomalously intense Raman spectra of pyridine at a
silver electrode. Journal of the American Chemical Society, 1977. 99(15): p. 5215-5217.
19. Otto, A., et al., Surface-enhanced Raman scattering. Journal of Physics: Condensed
Matter, 1992. 4(5): p. 1143.
20. Lakowicz, J.R., et al., Plasmon-controlled fluorescence: a new paradigm in fluorescence
spectroscopy. Analyst, 2008. 133(10): p. 1308-1346.
21. Johnson, P.B. and R.W. Christy, Optical Constants of the Noble Metals. Physical Review
B, 1972. 6(12): p. 4370-4379.
22. Willets, K.A. and R.P. Van Duyne, Localized Surface Plasmon Resonance Spectroscopy
and Sensing. Annual Review of Physical Chemistry, 2007. 58(1): p. 267-297.
23. Erwin, W.R., et al., Light trapping in mesoporous solar cells with plasmonic
nanostructures. Energy & Environmental Science, 2016. 9(5): p. 1577-1601.
24. Sharma, B., et al., SERS: Materials, applications, and the future. Materials Today, 2012.
15(1–2): p. 16-25.
25. Lee, K.-S. and M.A. El-Sayed, Gold and Silver Nanoparticles in Sensing and Imaging:
Sensitivity of Plasmon Response to Size, Shape, and Metal Composition. The Journal of
Physical Chemistry B, 2006. 110(39): p. 19220-19225.
26. Moskovits, M., Surface roughness and the enhanced intensity of Raman scattering by
molecules adsorbed on metals. The Journal of Chemical Physics, 1978. 69(9): p. 4159-
4161.
27. Ichinohe, Y., T. Wadayama, and A. Hatta, Electrochemical reduction of CO2 on silver as
probed by surface-enhanced Raman scattering. Journal of Raman Spectroscopy, 1995.
26(5): p. 335-340.
28. Lin, X.-F., et al., Surface-enhanced Raman spectroscopy with ultraviolet excitation.
Journal of Raman Spectroscopy, 2005. 36(6-7): p. 606-612.
29. Brolo, A.G., D.E. Irish, and B.D. Smith, Applications of surface enhanced Raman
scattering to the study of metal-adsorbate interactions. Journal of Molecular Structure,
1997. 405(1): p. 29-44.
30. Nickel, U., et al., A Silver Colloid Produced by Reduction with Hydrazine as Support for
Highly Sensitive Surface-Enhanced Raman Spectroscopy. Langmuir, 2000. 16(23): p.
9087-9091.
31. Turkevich, J., P.C. Stevenson, and J. Hillier, A study of the nucleation and growth
processes in the synthesis of colloidal gold. Discussions of the Faraday Society, 1951.
11(0): p. 55-75.
32. Frens, G., Controlled Nucleation for the Regulation of the Particle Size in Monodisperse
Gold Suspensions. Nature, 1973. 241: p. 20-22.
33. Hao, E., et al., Synthesis and Optical Properties of “Branched” Gold Nanocrystals. Nano
Letters, 2004. 4(2): p. 327-330.
34. Rodríguez-Lorenzo, L., et al., Zeptomol Detection Through Controlled Ultrasensitive
Surface-Enhanced Raman Scattering. Journal of the American Chemical Society, 2009.
131(13): p. 4616-4618.
35. Sönnichsen, C., et al., Drastic Reduction of Plasmon Damping in Gold Nanorods.
Physical Review Letters, 2002. 88(7): p. 077402.
36. Nehl, C.L., H. Liao, and J.H. Hafner, Optical Properties of Star-Shaped Gold
Nanoparticles. Nano Letters, 2006. 6(4): p. 683-688.
37. van der Zande, B.M.I., et al., Colloidal Dispersions of Gold Rods: Synthesis and Optical
Properties. Langmuir, 2000. 16(2): p. 451-458.
184

38. Schreiber, R., et al., Ordering Gold Nanoparticles with DNA Origami Nanoflowers. ACS
Nano, 2016. 10(8): p. 7303-7306.
39. Liu, N., et al., Three-Dimensional Plasmon Rulers. Science, 2011. 332(6036): p. 1407-
1410.
40. Ross, M.B., et al., Nanoscale form dictates mesoscale function in plasmonic DNA–
nanoparticle superlattices. Nat Nano, 2015. 10(5): p. 453-458.
41. Liz-Marzán, L.M., Tailoring Surface Plasmons through the Morphology and Assembly of
Metal Nanoparticles. Langmuir, 2006. 22(1): p. 32-41.
42. Halas, N.J., et al., Plasmons in Strongly Coupled Metallic Nanostructures. Chemical
Reviews, 2011. 111(6): p. 3913-3961.
43. Reinhard, B.M., et al., Calibration of Dynamic Molecular Rulers Based on Plasmon
Coupling between Gold Nanoparticles. Nano Letters, 2005. 5(11): p. 2246-2252.
44. Chung, T., et al., Plasmonic Nanostructures for Nano-Scale Bio-Sensing. Sensors, 2011.
11(11): p. 10907.
45. Su, K.H., et al., Interparticle Coupling Effects on Plasmon Resonances of Nanogold
Particles. Nano Letters, 2003. 3(8): p. 1087-1090.
46. Funston, A.M., et al., Plasmon Coupling of Gold Nanorods at Short Distances and in
Different Geometries. Nano Letters, 2009. 9(4): p. 1651-1658.
47. Oubre, C. and P. Nordlander, Finite-difference Time-domain Studies of the Optical
Properties of Nanoshell Dimers. The Journal of Physical Chemistry B, 2005. 109(20): p.
10042-10051.
48. Nordlander, P., et al., Plasmon Hybridization in Nanoparticle Dimers. Nano Letters,
2004. 4(5): p. 899-903.
49. Fromm, D.P., et al., Gap-Dependent Optical Coupling of Single “Bowtie” Nanoantennas
Resonant in the Visible. Nano Letters, 2004. 4(5): p. 957-961.
50. Atay, T., J.-H. Song, and A.V. Nurmikko, Strongly Interacting Plasmon Nanoparticle
Pairs: From Dipole−Dipole Interaction to Conductively Coupled Regime. Nano Letters,
2004. 4(9): p. 1627-1631.
51. Anker, J.N., et al., Biosensing with plasmonic nanosensors. Nat Mater, 2008. 7(6): p.
442-453.
52. Endo, T., et al., Multiple Label-Free Detection of Antigen−Antibody Reaction Using
Localized Surface Plasmon Resonance-Based Core−Shell Structured Nanoparticle Layer
Nanochip. Analytical Chemistry, 2006. 78(18): p. 6465-6475.
53. Nam, K.T., et al., Genetically Driven Assembly of Nanorings Based on the M13 Virus.
Nano Letters, 2003. 4(1): p. 23-27.
54. Bardhan, R., et al., Theranostic Nanoshells: From Probe Design to Imaging and
Treatment of Cancer. Accounts of Chemical Research, 2011. 44(10): p. 936-946.
55. Maier, S.A., et al., Local detection of electromagnetic energy transport below the
diffraction limit in metal nanoparticle plasmon waveguides. Nat Mater, 2003. 2(4): p.
229-232.
56. Ozbay, E., Plasmonics: Merging Photonics and Electronics at Nanoscale Dimensions.
Science, 2006. 311(5758): p. 189-193.
57. Cortie, M.B. and A.M. McDonagh, Synthesis and Optical Properties of Hybrid and Alloy
Plasmonic Nanoparticles. Chemical Reviews, 2011. 111(6): p. 3713-3735.
58. You, H., et al., Synthesis of colloidal metal and metal alloy nanoparticles for
electrochemical energy applications. Chemical Society Reviews, 2013. 42(7): p. 2880-
2904.
185

59. Katz, E. and I. Willner, Integrated Nanoparticle–Biomolecule Hybrid Systems: Synthesis,
Properties, and Applications. Angewandte Chemie International Edition, 2004. 43(45): p.
6042-6108.
60. Zavaleta, C.L., et al., Multiplexed imaging of surface enhanced Raman scattering
nanotags in living mice using noninvasive Raman spectroscopy. Proceedings of the
National Academy of Sciences, 2009. 106(32): p. 13511-13516.
61. Zavaleta, C., et al., Noninvasive Raman Spectroscopy in Living Mice for Evaluation of
Tumor Targeting with Carbon Nanotubes. Nano Letters, 2008. 8(9): p. 2800-2805.
62. Keren, S., et al., Noninvasive Molecular Imaging of Small Living Subjects Using Raman
Spectroscopy. Proceedings of the National Academy of Sciences, 2008. 105(15): p. 5844-
5849.
63. Wang, Y., et al., Biocompatibility and Biodistribution of Surface-Enhanced Raman
Scattering Nanoprobes in Zebrafish Embryos: In vivo and Multiplex Imaging. ACS
Nano, 2010. 4(7): p. 4039-4053.
64. Prencipe, G., et al., PEG Branched Polymer for Functionalization of Nanomaterials with
Ultralong Blood Circulation. Journal of the American Chemical Society, 2009. 131(13):
p. 4783-4787.
65. Fernández-López, C., et al., Highly Controlled Silica Coating of PEG-Capped Metal
Nanoparticles and Preparation of SERS-Encoded Particles. Langmuir, 2009. 25(24): p.
13894-13899.
66. Bearinger, J.P., et al., Chemisorbed poly(propylene sulphide)-based copolymers resist
biomolecular interactions. Nat Mater, 2003. 2(4): p. 259-264.
67. Rahme, K., et al., PEGylated gold nanoparticles: polymer quantification as a function of
PEG lengths and nanoparticle dimensions. RSC Advances, 2013. 3(17): p. 6085-6094.
68. Oh, E., et al., PEGylated Luminescent Gold Nanoclusters: Synthesis, Characterization,
Bioconjugation, and Application to One- and Two-Photon Cellular Imaging. Particle &
Particle Systems Characterization, 2013. 30(5): p. 453-466.
69. Zeng, S., et al., A Review on Functionalized Gold Nanoparticles for Biosensing
Applications. Plasmonics, 2011. 6(3): p. 491.
70. Park, J.-W. and J.S. Shumaker-Parry, Structural Study of Citrate Layers on Gold
Nanoparticles: Role of Intermolecular Interactions in Stabilizing Nanoparticles. Journal
of the American Chemical Society, 2014. 136(5): p. 1907-1921.
71. Sperling, R.A. and W.J. Parak, Surface modification, functionalization and
bioconjugation of colloidal inorganic nanoparticles. Philosophical Transactions of the
Royal Society A: Mathematical, Physical and
Engineering Sciences, 2010. 368(1915): p. 1333-1383.
72. Laaksonen, T., et al., Stability and Electrostatics of Mercaptoundecanoic Acid-Capped
Gold Nanoparticles with Varying Counterion Size. ChemPhysChem, 2006. 7(10): p.
2143-2149.
73. Howes, P.D., R. Chandrawati, and M.M. Stevens, Colloidal nanoparticles as advanced
biological sensors. Science, 2014. 346(6205).
74. Jadzinsky, P.D., et al., Structure of a Thiol Monolayer-Protected Gold Nanoparticle at
1.1 Å Resolution. Science, 2007. 318(5849): p. 430-433.
75. Ravindran, A., P. Chandran, and S.S. Khan, Biofunctionalized silver nanoparticles:
Advances and prospects. Colloids and Surfaces B: Biointerfaces, 2013. 105: p. 342-352.
76. Shang, L., et al., pH-Dependent Protein Conformational Changes in Albumin:Gold
Nanoparticle Bioconjugates: A Spectroscopic Study. Langmuir, 2007. 23(5): p. 2714-
2721.
186

77. Burt, J.L., et al., Noble-Metal Nanoparticles Directly Conjugated to Globular Proteins.
Langmuir, 2004. 20(26): p. 11778-11783.
78. Firestone, M.A., et al., Film Architecture in Biomolecular Assemblies. Effect of Linker on
the Orientation of Genetically Engineered Surface-Bound Proteins. Journal of the
American Chemical Society, 1996. 118(38): p. 9033-9041.
79. Lévy, R., et al., Rational and Combinatorial Design of Peptide Capping Ligands for Gold
Nanoparticles. Journal of the American Chemical Society, 2004. 126(32): p. 10076-
10084.
80. Mirkin, C.A., A DNA-based method for rationally assembling nanoparticles into
macroscopic materials. Nature, 1996. 382: p. 607-609.
81. Letsinger, R.L., et al., Use of a Steroid Cyclic Disulfide Anchor in Constructing Gold
Nanoparticle−Oligonucleotide Conjugates. Bioconjugate Chemistry, 2000. 11(2): p. 289-
291.
82. Ackerson, C.J., et al., Rigid, Specific, and Discrete Gold Nanoparticle/Antibody
Conjugates. Journal of the American Chemical Society, 2006. 128(8): p. 2635-2640.
83. Lee, J.M., et al., Direct Immobilization of Protein G Variants with Various Numbers of
Cysteine Residues on a Gold Surface. Analytical Chemistry, 2007. 79(7): p. 2680-2687.
84. Parak, W.J., et al., Conformation of Oligonucleotides Attached to Gold Nanocrystals
Probed by Gel Electrophoresis. Nano Letters, 2003. 3(1): p. 33-36.
85. Aggarwal, P., et al., Nanoparticle interaction with plasma proteins as it relates to
particle biodistribution, biocompatibility and therapeutic efficacy. Advanced Drug
Delivery Reviews, 2009. 61(6): p. 428-437.
86. Göppert, T.M. and R.H. Müller, Polysorbate-stabilized solid lipid nanoparticles as
colloidal carriers for intravenous targeting of drugs to the brain: Comparison of plasma
protein adsorption patterns. Journal of Drug Targeting, 2005. 13(3): p. 179-187.
87. Jokerst, J.V., et al., Nanoparticle PEGylation for imaging and therapy. Nanomedicine
(London, England), 2011. 6(4): p. 715-728.
88. Kanaras, A.G., et al., Thioalkylated tetraethylene glycol: a new ligand for water soluble
monolayer protected gold clusters. Chemical Communications, 2002(20): p. 2294-2295.
89. Schoenfisch, M.H. and J.E. Pemberton, Air Stability of Alkanethiol Self-Assembled
Monolayers on Silver and Gold Surfaces. Journal of the American Chemical Society,
1998. 120(18): p. 4502-4513.
90. Garrell, R.L., et al., Adsorption of Sulfur Containing Molecules on Gold: The Effect of
Oxidation on Monolayer Formation and Stability Characterized by Experiments and
Theory. Journal of the American Chemical Society, 1995. 117(46): p. 11563-11571.
91. Le Ru, E.C., et al., Surface Enhanced Raman Scattering Enhancement Factors: A
Comprehensive Study. The Journal of Physical Chemistry C, 2007. 111(37): p. 13794-
13803.
92. McNay, G., et al., Surface-Enhanced Raman Scattering (SERS) and Surface-Enhanced
Resonance Raman Scattering (SERRS): A Review of Applications. Applied Spectroscopy,
2011. 65(8): p. 825-837.
93. Qian, X., et al., In vivo Tumor Targeting and Spectroscopic Detection with Surface-
Enhanced Raman Nanoparticle Tags. Nat Biotech, 2008. 26(1): p. 83-90.
94. Samanta, A., et al., Ultrasensitive Near-Infrared Raman Reporters for SERS-Based In
Vivo Cancer Detection. Angewandte Chemie International Edition, 2011. 50(27): p.
6089-6092.
95. Wang, Y., B. Yan, and L. Chen, SERS Tags: Novel Optical Nanoprobes for Bioanalysis.
Chemical Reviews, 2013. 113(3): p. 1391-1428.
187

96. Jun, B.-H., et al., Surface-Enhanced Raman Spectroscopic-Encoded Beads for Multiplex
Immunoassay. Journal of Combinatorial Chemistry, 2007. 9(2): p. 237-244.
97. Matschulat, A., D. Drescher, and J. Kneipp, Surface-Enhanced Raman Scattering Hybrid
Nanoprobe Multiplexing and Imaging in Biological Systems. ACS Nano, 2010. 4(6): p.
3259-3269.
98. Cody, C.W., et al., Chemical structure of the hexapeptide chromophore of the Aequorea
green-fluorescent protein. Biochemistry, 1993. 32(5): p. 1212-1218.
99. Brejc, K., et al., Structural basis for dual excitation and photoisomerization of the
Aequorea victoria green fluorescent protein. Proceedings of the National Academy of
Sciences, 1997. 94(6): p. 2306-2311.
100. Habuchi, S., et al., Single-Molecule Surface Enhanced Resonance Raman Spectroscopy
of the Enhanced Green Fluorescent Protein. Journal of the American Chemical Society,
2003. 125(28): p. 8446-8447.
101. Heim, R. and R.Y. Tsien, Engineering green fluorescent protein for improved brightness,
longer wavelengths and fluorescence resonance energy transfer. Current Biology, 1996.
6(2): p. 178-182.
102. Ip, S., et al., Phospholipid Membrane Encapsulation of Nanoparticles for Surface-
Enhanced Raman Scattering. Langmuir, 2011. 27(11): p. 7024-7033.
103. Huang, C.-C., et al., Synthesis of silica-coated gold nanorod as Raman tags by
modulating cetyltrimethylammonium bromide concentration. Colloids and Surfaces A:
Physicochemical and Engineering Aspects, 2012. 409: p. 61-68.
104. Sonnichsen, C., et al., A molecular ruler based on plasmon coupling of single gold and
silver nanoparticles. Nat Biotech, 2005. 23(6): p. 741-745.
105. Gersten, J. and A. Nitzan, Electromagnetic theory of enhanced Raman scattering by
molecules adsorbed on rough surfaces. The Journal of Chemical Physics, 1980. 73(7): p.
3023-3037.
106. H Metiu, a. and P. Das, The Electromagnetic Theory of Surface Enhanced Spectroscopy.
Annual Review of Physical Chemistry, 1984. 35(1): p. 507-536.
107. Camden, J.P., et al., Probing the Structure of Single-Molecule Surface-Enhanced Raman
Scattering Hot Spots. Journal of the American Chemical Society, 2008. 130(38): p.
12616-12617.
108. Kneipp, K., et al., Single Molecule Detection Using Surface-Enhanced Raman Scattering
(SERS). Physical Review Letters, 1997. 78(9): p. 1667-1670.
109. Wustholz, K.L., et al., Structure−Activity Relationships in Gold Nanoparticle Dimers
and Trimers for Surface-Enhanced Raman Spectroscopy. Journal of the American
Chemical Society, 2010. 132(31): p. 10903-10910.
110. Lim, D.-K., et al., Highly uniform and reproducible surface-enhanced Raman scattering
from DNA-tailorable nanoparticles with 1-nm interior gap. Nat Nano, 2011. 6(7): p. 452-
460.
111. Tsien, R.Y., THE GREEN FLUORESCENT PROTEIN. Annual Review of Biochemistry,
1998. 67(1): p. 509-544.
112. Bell, A.F., et al., Probing the Ground State Structure of the Green Fluorescent Protein
Chromophore Using Raman Spectroscopyâ€ Biochemistry, 2000. 39(15): p. 4423-4431.
113. He, X., A.F. Bell, and P.J. Tonge, Isotopic Labeling and Normal-Mode Analysis of a
Model Green Fluorescent Protein Chromophore. The Journal of Physical Chemistry B,
2002. 106(23): p. 6056-6066.
114. Danckwerts, M. and L. Novotny, Optical Frequency Mixing at Coupled Gold
Nanoparticles. Physical Review Letters, 2007. 98(2): p. 026104.
188

115. Kabashin, A.V., et al., Plasmonic nanorod metamaterials for biosensing. Nat Mater,
2009. 8(11): p. 867-871.
116. Cole, J.R., et al., Photothermal Efficiencies of Nanoshells and Nanorods for Clinical
Therapeutic Applications. The Journal of Physical Chemistry C, 2009. 113(28): p. 12090-
12094.
117. Kang, Y., et al., Plasmonic Hot Electron Induced Structural Phase Transition in a MoS2
Monolayer. Advanced Materials, 2014. 26(37): p. 6467-6471.
118. Taton, T.A., C.A. Mirkin, and R.L. Letsinger, Scanometric DNA Array Detection with
Nanoparticle Probes. Science, 2000. 289(5485): p. 1757-1760.
119. Cao, Y.C., R. Jin, and C.A. Mirkin, Nanoparticles with Raman Spectroscopic
Fingerprints for DNA and RNA Detection. Science, 2002. 297(5586): p. 1536-1540.
120. Bohndiek, S.E., et al., A small animal Raman instrument for rapid, wide-area,
spectroscopic imaging. Proceedings of the National Academy of Sciences, 2013.
110(30): p. 12408-12413.
121. Ayala-Orozco, C., et al., Au Nanomatryoshkas as Efficient Near-Infrared Photothermal
Transducers for Cancer Treatment: Benchmarking against Nanoshells. ACS Nano, 2014.
8(6): p. 6372-6381.
122. Winter, J., et al., Many roads to maturity: microRNA biogenesis pathways and their
regulation. Nat Cell Biol, 2009. 11(3): p. 228-234.
123. Esquela-Kerscher, A. and F.J. Slack, Oncomirs [mdash] microRNAs with a role in
cancer. Nat Rev Cancer, 2006. 6(4): p. 259-269.
124. Joshi, G.K., et al., Highly Specific Plasmonic Biosensors for Ultrasensitive MicroRNA
Detection in Plasma from Pancreatic Cancer Patients. Nano Letters, 2014. 14(12): p.
6955-6963.
125. La Rocca, R., et al., Out-of-Plane Plasmonic Antennas for Raman Analysis in Living
Cells. Small, 2015. 11(36): p. 4632-4637.
126. Yi, N., et al., A hybrid system with highly enhanced graphene SERS for rapid and tag-
free tumor cells detection. Scientific Reports, 2016. 6: p. 25134.
127. Kircher, M.F., et al., A Brain Tumor Molecular Imaging Strategy Using A New Triple-
Modality MRI-Photoacoustic-Raman Nanoparticle. Nature medicine, 2012. 18(5): p.
829-834.
128. De La Zerda, A., et al., Carbon nanotubes as photoacoustic molecular imaging agents in
living mice. Nat Nano, 2008. 3(9): p. 557-562.
129. de la Zerda, A., et al., Advanced contrast nanoagents for photoacoustic molecular
imaging, cytometry, blood test and photothermal theranostics. Contrast Media &
Molecular Imaging, 2011. 6(5): p. 346-369.
130. Wang, X., et al., Noninvasive laser-induced photoacoustic tomography for structural and
functional in vivo imaging of the brain. Nat Biotech, 2003. 21(7): p. 803-
806.
131. Toi, M., et al., Visualization of tumor-related blood vessels in human breast by
photoacoustic imaging system with a hemispherical detector array. Scientific Reports,
2017. 7: p. 41970.
132. Favazza, C.P., L.A. Cornelius, and L.V. Wang, In vivo functional photoacoustic
microscopy of cutaneous microvasculature in human skin. Journal of Biomedical Optics,
2011. 16(2): p. 026004.
133. Wang, L.V., Multiscale photoacoustic microscopy and computed tomography. Nat
Photon, 2009. 3(9): p. 503-509.
189

134. Xia, J., J. Yao, and L.V. Wang, Photoacoustic tomography: principles and advances.
Electromagnetic waves (Cambridge, Mass.), 2014. 147: p. 1-22.
135. Beard, P., Biomedical photoacoustic imaging. Interface Focus, 2011. 1(4): p. 602-631.
136. Pramanik, M. and L.V. Wang, Thermoacoustic and photoacoustic sensing of
temperature. Journal of Biomedical Optics, 2009. 14(5): p. 054024-054024-7.
137. Wang, L.V. and S. Hu, Photoacoustic Tomography: In Vivo Imaging from Organelles to
Organs. Science, 2012. 335(6075): p. 1458-1462.
138. Ku, G., et al., Imaging of tumor angiogenesis in rat brains in vivo by photoacoustic
tomography. Applied Optics, 2005. 44(5): p. 770-775.
139. Frangioni, J.V., New Technologies for Human Cancer Imaging. Journal of Clinical
Oncology, 2008. 26(24): p. 4012-4021.
140. Huang, X., et al., Cancer Cell Imaging and Photothermal Therapy in the Near-Infrared
Region by Using Gold Nanorods. Journal of the American Chemical Society, 2006.
128(6): p. 2115-2120.
141. Yang, X., et al., Photoacoustic Tomography of a Rat Cerebral Cortex in vivo with Au
Nanocages as an Optical Contrast Agent. Nano Letters, 2007. 7(12): p. 3798-3802.
142. Kim, D., Y.Y. Jeong, and S. Jon, A Drug-Loaded Aptamer−Gold Nanoparticle
Bioconjugate for Combined CT Imaging and Therapy of Prostate Cancer. ACS Nano,
2010. 4(7): p. 3689-3696.
143. Schipper, M.L., et al., A pilot toxicology study of single-walled carbon nanotubes in a
small sample of mice. Nat Nano, 2008. 3(4): p. 216-221.
144. Kam, N.W.S., et al., Carbon nanotubes as multifunctional biological transporters and
near-infrared agents for selective cancer cell destruction. Proceedings of the National
Academy of Sciences of the United States of America, 2005. 102(33): p. 11600-11605.
145. Lam, C.-W., et al., Pulmonary Toxicity of Single-Wall Carbon Nanotubes in Mice 7 and
90 Days After Intratracheal Instillation. Toxicological Sciences, 2004. 77(1): p. 126-134.
146. Warheit, D.B., et al., Comparative Pulmonary Toxicity Assessment of Single-wall Carbon
Nanotubes in Rats. Toxicological Sciences, 2004. 77(1): p. 117-125.
147. Kim, S.T., et al., The Role of Surface Functionality in Determining Nanoparticle
Cytotoxicity. Accounts of Chemical Research, 2013. 46(3): p. 681-691.
148. Tian, C., et al., Plasmonic Nanoparticles with Quantitatively Controlled Bioconjugation
for Photoacoustic Imaging of Live Cancer Cells. Advanced Science, 2016. 3(12): p.
1600237-n/a.
149. Hamada, K., et al., Raman microscopy for dynamic molecular imaging of living cells.
Journal of Biomedical Optics, 2008. 13(4): p. 044027-044027-4.
150. Marcet, S., et al. Raman spectroscopy hyperspectral imager based on Bragg tunable
filters. 2012.
151. Kang, J.W., et al., High Resolution Live Cell Raman Imaging Using Subcellular
Organelle-Targeting SERS-Sensitive Gold Nanoparticles with Highly Narrow Intra-
Nanogap. Nano Letters, 2015. 15(3): p. 1766-1772.
152. Williams, K.P.J., et al., Use of a rapid scanning stigmatic Raman imaging spectrograph
in the industrial environment. Journal of Raman Spectroscopy, 1994. 25(1): p. 131-138.
153. Aroca, R.F. and C.J.L. Constantino, Surface-Enhanced Raman Scattering: Imaging and
Mapping of Langmuir−Blodgett Monolayers Physically Adsorbed onto Silver Island
Films. Langmuir, 2000. 16(12): p. 5425-5429.
154. Constantino, C., J. Duff, and R. Aroca, Surface enhanced resonance Raman scattering
imaging of Langmuir–Blodgett monolayers of bis (benzimidazo) thioperylene.
190

Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2001. 57(6): p.
1249-1259.
155. GaufresE, et al., Giant Raman scattering from J-aggregated dyes inside carbon
nanotubes for multispectral imaging. Nat Photon, 2014. 8(1): p. 72-78.
156. Boisselier, E. and D. Astruc, Gold nanoparticles in nanomedicine: preparations,
imaging, diagnostics, therapies and toxicity. Chemical society reviews, 2009. 38(6): p.
1759-1782.
157. Nel, A.E., et al., Understanding biophysicochemical interactions at the nano-bio
interface. Nat Mater, 2009. 8(7): p. 543-557.
158. Qiu, L., et al., A Cell-Targeted, Size-Photocontrollable, Nuclear-Uptake Nanodrug
Delivery System for Drug-Resistant Cancer Therapy. Nano Letters, 2015. 15(1): p. 457-
463.
159. Ding, Y., et al., The performance of thiol-terminated PEG-paclitaxel-conjugated gold
nanoparticles. Biomaterials, 2013. 34(38): p. 10217-10227.
160. Hirsch, L.R., et al., Nanoshell-mediated near-infrared thermal therapy of tumors under
magnetic resonance guidance. Proceedings of the National Academy of Sciences, 2003.
100(23): p. 13549-13554.
161. Weissleder, R., A clearer vision for in vivo imaging. Nat Biotech, 2001. 19(4): p. 316-
317.
162. Cherukuri, P., E.S. Glazer, and S.A. Curley, Targeted hyperthermia using metal
nanoparticles. Advanced Drug Delivery Reviews, 2010. 62(3): p. 339-345.
163. Huang, H.-C., K. Rege, and J.J. Heys, Spatiotemporal Temperature Distribution and
Cancer Cell Death in Response to Extracellular Hyperthermia Induced by Gold
Nanorods. ACS Nano, 2010. 4(5): p. 2892-2900.
164. von Maltzahn, G., et al., SERS-Coded Gold Nanorods as a Multifunctional Platform for
Densely Multiplexed Near-Infrared Imaging and Photothermal Heating. Advanced
Materials, 2009. 21(31): p. 3175-3180.
165. Chen, M., et al., Core–Shell Pd@Au Nanoplates as Theranostic Agents for In-Vivo
Photoacoustic Imaging, CT Imaging, and Photothermal Therapy. Advanced Materials,
2014. 26(48): p. 8210-8216.
166. Huang, P., et al., Folic acid-conjugated Silica-modified gold nanorods for X-ray/CT
imaging-guided dual-mode radiation and photo-thermal therapy. Biomaterials, 2011.
32(36): p. 9796-9809.
167. Kim, Y.E., et al., Green fluorescent protein nanopolygons as monodisperse
supramolecular assemblies of functional proteins with defined valency. Nature
Communications, 2015. 6: p. 7134.
168. Pinaud, F. and M. Dahan, Targeting and imaging single biomolecules in living cells by
complementation-activated light microscopy with split-fluorescent proteins. Proceedings
of the National Academy of Sciences, 2011. 108(24): p. E201-E210.
169. Feinberg, E.H., et al., GFP Reconstitution Across Synaptic Partners (GRASP) Defines
Cell Contacts and Synapses in Living Nervous Systems. Neuron, 2008. 57(3): p. 353-363.
170. Prasher, D.C., et al., Primary structure of the Aequorea victoria green-fluorescent
protein. Gene, 1992. 111(2): p. 229-233.
171. Chalfie, M., et al., Green fluorescent protein as a marker for gene expression. Science,
1994. 263(5148): p. 802-805.
172. Heim, R., D.C. Prasher, and R.Y. Tsien, Wavelength mutations and posttranslational
autoxidation of green fluorescent protein. Proceedings of the National Academy of
Sciences of the United States of America, 1994. 91(26): p. 12501-12504.
191

173. Shimomura, O., F.H. Johnson, and Y. Saiga, Extraction, Purification and Properties of
Aequorin, a Bioluminescent Protein from the Luminous Hydromedusan, Aequorea.
Journal of Cellular and Comparative Physiology, 1962. 59(3): p. 223-239.
174. Zhou, X.X. and M.Z. Lin, Photoswitchable fluorescent proteins: ten years of colorful
chemistry and exciting applications. Current Opinion in Chemical Biology, 2013. 17(4):
p. 682-690.
175. Cabantous, S., T.C. Terwilliger, and G.S. Waldo, Protein tagging and detection with
engineered self-assembling fragments of green fluorescent protein. Nat Biotech, 2005.
23(1): p. 102-107.
176. Ormo, M., et al., Crystal structure of the Aequorea victoria green fluorescent protein.
Science, 1996. 273(5280): p. 1392.
177. Heim, R., A.B. Cubitt, and R.Y. Tsien, Improved green fluorescence. Nature, 1995.
373(6516): p. 663-664.
178. Pedelacq, J.-D., et al., Engineering and characterization of a superfolder green
fluorescent protein. Nat Biotech, 2006. 24(1): p. 79-88.
179. Ward, W.W. and S.H. Bokman, Reversible denaturation of Aequorea green-fluorescent
protein: physical separation and characterization of the renatured protein. Biochemistry,
1982. 21(19): p. 4535-4540.
180. Reid, B.G. and G.C. Flynn, Chromophore Formation in Green Fluorescent Protein.
Biochemistry, 1997. 36(22): p. 6786-6791.
181. Iizuka, R., M. Yamagishi-Shirasaki, and T. Funatsu, Kinetic study of de novo
chromophore maturation of fluorescent proteins. Analytical Biochemistry, 2011. 414(2):
p. 173-178.
182. Wachter, R.M., et al., Structural basis of spectral shifts in the yellow-emission variants of
green fluorescent protein. Structure, 1998. 6(10): p. 1267-1277.
183. Zhong, S., D. Navaratnam, and J. Santos-Sacchi, A Genetically-Encoded YFP Sensor
with Enhanced Chloride Sensitivity, Photostability and Reduced pH Interference
Demonstrates Augmented Transmembrane Chloride Movement by Gerbil Prestin
(SLC26a5). PLOS ONE, 2014. 9(6): p. e99095.
184. Lockard, M.A. and G.S. Waldo, Cyan and yellow fluorescent color variants of split gfp.
2012, Google Patents.
185. Shu, X., et al., Novel Chromophores and Buried Charges Control Color in mFruits.
Biochemistry, 2006. 45(32): p. 9639-9647.
186. Lelimousin, M., et al., Intrinsic Dynamics in ECFP and Cerulean Control Fluorescence
Quantum Yield. Biochemistry, 2009. 48(42): p. 10038-10046.
187. Fisher, A.C. and M.P. DeLisa, Laboratory Evolution of Fast-Folding Green Fluorescent
Protein Using Secretory Pathway Quality Control. PLOS ONE, 2008. 3(6): p. e2351.
188. Toseland, C.P., Fluorescent labeling and modification of proteins. Journal of Chemical
Biology, 2013. 6(3): p. 85-95.
189. Suhling, K., et al., Imaging the Environment of Green Fluorescent Protein. Biophysical
Journal, 2002. 83(6): p. 3589-3595.
190. Lakowicz, J.R., Principles of Fluorescence Spectroscopy. 3 ed. 2006, US: Springer.
191. Berezin, M.Y. and S. Achilefu, Fluorescence Lifetime Measurements and Biological
Imaging. Chemical reviews, 2010. 110(5): p. 2641-2684.
192. Chattoraj, M., et al., Ultra-fast excited state dynamics in green fluorescent protein:
multiple states and proton transfer. Proceedings of the National Academy of Sciences,
1996. 93(16): p. 8362-8367.
192

193. Striker, G., et al., Photochromicity and Fluorescence Lifetimes of Green Fluorescent
Protein. The Journal of Physical Chemistry B, 1999. 103(40): p. 8612-8617.
194. Jung, G., J. Wiehler, and A. Zumbusch, The Photophysics of Green Fluorescent Protein:
Influence of the Key Amino Acids at Positions 65, 203, and 222. Biophysical Journal,
2005. 88(3): p. 1932-1947.
195. Wineman-Fisher, V., et al., Insight into the structure and the mechanism of the slow
proton transfer in the GFP double mutant T203V/S205A. Physical Chemistry Chemical
Physics, 2014. 16(23): p. 11211-11223.
196. Creemers, T.M.H., et al., Three photoconvertible forms of green fluorescent protein
identified by spectral hole-burning. Nat Struct Mol Biol, 1999. 6(6): p. 557-560.
197. Hyun Bae, J., et al., Expansion of the Genetic Code Enables Design of a Novel “Gold”
Class of Green Fluorescent Proteins. Journal of Molecular Biology, 2003. 328(5): p.
1071-1081.
198. Jung, G., et al., Fluorescence Lifetime of Fluorescent Proteins, in Fluorescent Proteins I:
From Understanding to Design, G. Jung, Editor. 2012, Springer Berlin Heidelberg:
Berlin, Heidelberg. p. 69-97.
199. Villoing, A., et al., Complex Fluorescence of the Cyan Fluorescent Protein: Comparisons
with the H148D Variant and Consequences for Quantitative Cell Imaging. Biochemistry,
2008. 47(47): p. 12483-12492.
200. Rizzo, M.A., et al., An improved cyan fluorescent protein variant useful for FRET. Nat
Biotech, 2004. 22(4): p. 445-449.
201. Würth, C., et al., Relative and absolute determination of fluorescence quantum yields of
transparent samples. Nat. Protocols, 2013. 8(8): p. 1535-1550.
202. Weber, G. and F.W.J. Teale, Determination of the absolute quantum yield of fluorescent
solutions. Transactions of the Faraday Society, 1957. 53(0): p. 646-655.
203. Cubitt, A.B., L.A. Woollenweber, and R. Heim, Chapter 2: Understanding Structure—
Function Relationships in the Aequorea victoria Green Fluorescent Protein. Methods in
Cell Biology, 1998. 58: p. 19-30.
204. Shaner, N.C., et al., A bright monomeric green fluorescent protein derived from
Branchiostoma lanceolatum. Nat Meth, 2013. 10(5): p. 407-409.
205. Sinnecker, D., et al., Reversible Photobleaching of Enhanced Green Fluorescent
Proteinsâ€ Biochemistry, 2005. 44(18): p. 7085-7094.
206. Dickson, R.M., et al., On/off blinking and switching behaviour of single molecules of
green fluorescent protein. Nature, 1997. 388(6640): p. 355-358.
207. Weissman, J.S., et al., Characterization of the Active Intermediate of a GroEL–GroES-
Mediated Protein Folding Reaction. Cell, 1996. 84(3): p. 481-490.
208. Ward, W.W., et al., SPECTROPHOTOMETRIC IDENTITY OF THE ENERGY
TRANSFER CHROMOPHORES IN RENILLA AND AEQUOREA GREEN-
FLUORESCENT PROTEINS. Photochemistry and Photobiology, 1980. 31(6): p. 611-
615.
209. Kremers, G.-J., et al., Cyan and Yellow Super Fluorescent Proteins with Improved
Brightness, Protein Folding, and FRET Förster Radius. Biochemistry, 2006. 45(21): p.
6570-6580.
210. Adams, S.R., et al., New biarsenical Ligands and tetracysteine motifs for protein labeling
in vitro and in vivo: Synthesis and biological applications. Journal of the American
Chemical Society, 2002. 124(21): p. 6063-6076.
211. Gradinaru, C.C., et al., Fluorescence anisotropy: from single molecules to live cells.
Analyst, 2010. 135(3): p. 452-459.
193

212. Schlosshauer, M. and D. Baker, Realistic protein–protein association rates from a simple
diffusional model neglecting long-range interactions, free energy barriers, and landscape
ruggedness. Protein Science : A Publication of the Protein Society, 2004. 13(6): p. 1660-
1669.
213. Paladino, S., et al., Protein oligomerization modulates raft partitioning and apical sorting
of GPI-anchored proteins. The Journal of Cell Biology, 2004. 167(4): p. 699-709.
214. Friedrichson, T. and T.V. Kurzchalia, Microdomains of GPI-anchored proteins in living
cells revealed by crosslinking. Nature, 1998. 394(6695): p. 802-805.
215. Serge, A., et al., Dynamic multiple-target tracing to probe spatiotemporal cartography of
cell membranes. Nat Meth, 2008. 5(8): p. 687-694.
216. Thompson, R.E., D.R. Larson, and W.W. Webb, Precise nanometer localization analysis
for individual fluorescent probes. Biophysical Journal, 2002. 82(5): p. 2775-2783.
217. Henglein, A., Small-particle research: physicochemical properties of extremely small
colloidal metal and semiconductor particles. Chemical Reviews, 1989. 89(8): p. 1861-
1873.
218. Jain, P.K., et al., Noble Metals on the Nanoscale: Optical and Photothermal Properties
and Some Applications in Imaging, Sensing, Biology, and Medicine. Accounts of
Chemical Research, 2008. 41(12): p. 1578-1586.
219. Penn, S.G., L. He, and M.J. Natan, Nanoparticles for bioanalysis. Current Opinion in
Chemical Biology, 2003. 7(5): p. 609-615.
220. Yang, X., et al., Nanoparticles for photoacoustic imaging. Wiley Interdisciplinary
Reviews: Nanomedicine and Nanobiotechnology, 2009. 1(4): p. 360-368.
221. Qian, X.M. and S.M. Nie, Single-molecule and single-nanoparticle SERS: from
fundamental mechanisms to biomedical applications. Chemical Society Reviews, 2008.
37(5): p. 912-920.
222. Aroca, R.F., et al., Surface-enhanced Raman scattering on colloidal nanostructures.
Advances in Colloid and Interface Science, 2005. 116(1): p. 45-61.
223. Kleinman, S.L., et al., Creating, characterizing, and controlling chemistry with SERS hot
spots. Physical Chemistry Chemical Physics, 2013. 15(1): p. 21-36.
224. Fabris, L., SERS Tags: The Next Promising Tool for Personalized Cancer Detection?
ChemNanoMat, 2016. 2(4): p. 249-258.
225. Wang, Y. and S. Schlucker, Rational design and synthesis of SERS labels. Analyst, 2013.
138(8): p. 2224-2238.
226. Abalde-Cela, S., et al., Surface-enhanced Raman scattering biomedical applications of
plasmonic colloidal particles. Journal of The Royal Society Interface, 2010.
227. Jimenez de Aberasturi, D., et al., Surface Enhanced Raman Scattering Encoded Gold
Nanostars for Multiplexed Cell Discrimination. Chemistry of Materials, 2016. 28(18): p.
6779-6790.
228. Kim, J.-H., et al., Nanoparticle Probes with Surface Enhanced Raman Spectroscopic
Tags for Cellular Cancer Targeting. Analytical Chemistry, 2006. 78(19): p. 6967-6973.
229. Romo-Herrera, J.M., R.A. Alvarez-Puebla, and L.M. Liz-Marzan, Controlled assembly of
plasmonic colloidal nanoparticle clusters. Nanoscale, 2011. 3(4): p. 1304-1315.
230. Le Ru, E.C. and P.G. Etchegoin, Single-molecule surface-enhanced Raman spectroscopy.
Annual review of physical chemistry, 2012. 63: p. 65-87.
231. Jiang, et al., Single Molecule Raman Spectroscopy at the Junctions of Large Ag
Nanocrystals. The Journal of Physical Chemistry B, 2003. 107(37): p. 9964-9972.
194

232. Futamata, M., Y. Maruyama, and M. Ishikawa, Critical importance of the junction in
touching Ag particles for single molecule sensitivity in SERS. Journal of Molecular
Structure, 2005. 735: p. 75-84.
233. Nam, J.-M., et al., Plasmonic Nanogap-Enhanced Raman Scattering with Nanoparticles.
Accounts of Chemical Research, 2016. 49(12): p. 2746-2755.
234. Natan, M.J., Concluding Remarks Surface enhanced Raman scattering. Faraday
Discussions, 2006. 132(0): p. 321-328.
235. Hao, E. and G.C. Schatz, Electromagnetic fields around silver nanoparticles and dimers.
The Journal of Chemical Physics, 2004. 120(1): p. 357-366.
236. Zhu, W. and K.B. Crozier, Quantum mechanical limit to plasmonic enhancement as
observed by surface-enhanced Raman scattering. 2014. 5: p. 5228.
237. Radziuk, D. and H. Moehwald, Prospects for plasmonic hot spots in single molecule
SERS towards the chemical imaging of live cells. Physical Chemistry Chemical Physics,
2015. 17(33): p. 21072-21093.
238. Taylor, R.W., et al., Precise Subnanometer Plasmonic Junctions for SERS within Gold
Nanoparticle Assemblies Using Cucurbit[n]uril “Glue”. ACS Nano, 2011. 5(5): p. 3878-
3887.
239. Fabris, L., Bottom-up optimization of SERS hot-spots. Chemical Communications, 2012.
48(75): p. 9346-9348.
240. Qian, X., X. Zhou, and S. Nie, Surface-Enhanced Raman Nanoparticle Beacons Based
on Bioconjugated Gold Nanocrystals and Long Range Plasmonic Coupling. Journal of
the American Chemical Society, 2008. 130(45): p. 14934-14935.
241. Graham, D., et al., Control of enhanced Raman scattering using a DNA-based assembly
process of dye-coded nanoparticles. Nat Nano, 2008. 3(9): p. 548-551.
242. Lim, D.-K., et al., Nanogap-engineerable Raman-active nanodumbbells for single-
molecule detection. Nat Mater, 2010. 9(1): p. 60-67.
243. Claridge, S.A., et al., Isolation of Discrete Nanoparticle−DNA Conjugates for Plasmonic
Applications. Nano Letters, 2008. 8(4): p. 1202-1206.
244. Koshland, D.E., The Key–Lock Theory and the Induced Fit Theory. Angewandte Chemie
International Edition in English, 1995. 33(23-24): p. 2375-2378.
245. Zhan, H., et al., In vivo single-molecule imaging identifies altered dynamics of calcium
channels in dystrophin-mutant C. elegans. Nat Commun. 5.
246. Bale, S.S., et al., A GFP complementation system for monitoring and directing
nanomaterial mediated protein delivery to human cellular organelles. Biotechnology and
Bioengineering, 2010. 107(6): p. 1040-1047.
247. Reed, A.M.W. and S.J. Metallo, Oriented Protein Adsorption to Gold Nanoparticles
through a Genetically Encodable Binding Motif. Langmuir, 2010. 26(24): p. 18945-
18950.
248. Sandroff, C.J. and D.R. Herschbach, Surface-enhanced Raman study of organic sulfides
adsorbed on silver: facile cleavage of sulfur-sulfur and carbon-sulfur bonds. The Journal
of Physical Chemistry, 1982. 86(17): p. 3277-3279.
249. Mock, J.J., D.R. Smith, and S. Schultz, Local Refractive Index Dependence of Plasmon
Resonance Spectra from Individual Nanoparticles. Nano Letters, 2003. 3(4): p. 485-491.
250. Weitz, D.A. and M.Y. Lin, Dynamic Scaling of Cluster-Mass Distributions in Kinetic
Colloid Aggregation. Physical Review Letters, 1986. 57(16): p. 2037-2040.
251. Lin, M.Y., et al., Universality in colloid aggregation. Nature, 1989. 339(6223): p. 360-
362.
195

252. Yang, F., L.G. Moss, and G.N. Phillips, The molecular structure of green fluorescent
protein. Nat Biotech, 1996. 14(10): p. 1246-1251.
253. Chung, T., T. Koker, and F. Pinaud, Split-GFP: SERS Enhancers in Plasmonic
Nanocluster Probes. Small, 2016. 12(42): p. 5891-5901.
254. Podstawka, E., et al., Food additives characterization by infrared, Raman, and surface-
enhanced Raman spectroscopies. Journal of Raman Spectroscopy, 2007. 38(3): p. 356-
363.
255. Asher, S.A., A.V. Mikhonin, and S. Bykov, UV Raman Demonstrates that α-Helical
Polyalanine Peptides Melt to Polyproline II Conformations. Journal of the American
Chemical Society, 2004. 126(27): p. 8433-8440.
256. Habuchi, S. and J. Hofkens, Single-Molecule Surface-Enhanced Resonance Raman
Spectroscopy of the Enhanced Green Fluorescent Protein EGFP, in Surface-Enhanced
Raman Scattering: Physics and Applications, K. Kneipp, M. Moskovits, and H. Kneipp,
Editors. 2006, Springer Berlin Heidelberg: Berlin, Heidelberg. p. 297-312.
257. Pinaud, F., et al., Dynamic Partitioning of a Glycosyl-Phosphatidylinositol-Anchored
Protein in Glycosphingolipid-Rich Microdomains Imaged by Single-Quantum Dot
Tracking. Traffic, 2009. 10(6): p. 691-712.
258. Yong Wu, X.W., Ligia Toro, Enrico Stefani, Highly Nonlinear Luminescence Induced by
Gold Nanoparticles on Glass Surfaces with Continuous-Wave Laser Illumination. 2015.
259. Dulkeith, E., et al., Plasmon emission in photoexcited gold nanoparticles. Physical
Review B, 2004. 70(20): p. 205424.
260. Wang, Z.B., et al., The influences of particle number on hot spots in strongly coupled
metal nanoparticles chain. The Journal of Chemical Physics, 2008. 128(9): p. 094705.
261. van der Wijk, T., et al., Increased Vesicle Recycling in Response to Osmotic Cell
Swelling: CAUSE AND CONSEQUENCE OF HYPOTONICITY-PROVOKED ATP
RELEASE. Journal of Biological Chemistry, 2003. 278(41): p. 40020-40025.
262. Yamashiro, D.J. and F.R. Maxfield, Kinetics of endosome acidification in mutant and
wild-type Chinese hamster ovary cells. The Journal of Cell Biology, 1987. 105(6): p.
2713-2721.
263. Mallidi, S., et al., Molecular specific optoacoustic imaging with plasmonic nanoparticles.
Optics Express, 2007. 15(11): p. 6583-6588.
264. Li, W. and X. Chen, Gold nanoparticles for photoacoustic imaging. Nanomedicine
(London, England), 2015. 10(2): p. 299-320.
265. Bayer, C.L., et al., Photoacoustic signal amplification through plasmonic nanoparticle
aggregation. Journal of Biomedical Optics, 2013. 18(1): p. 016001.
266. Evans, W., et al., Effect of aggregation and interfacial thermal resistance on thermal
conductivity of nanocomposites and colloidal nanofluids. International Journal of Heat
and Mass Transfer, 2008. 51(5): p. 1431-1438.
267. Nam, S.Y., et al., Nonlinear photoacoustic signal increase from endocytosis of gold
nanoparticles. Optics Letters, 2012. 37(22): p. 4708-4710.
268. Rodriguez, E.A., et al., The Growing and Glowing Toolbox of Fluorescent and
Photoactive Proteins. Trends in Biochemical Sciences. 42(2): p. 111-129.
269. Kent, K.P. and S.G. Boxer, Light-Activated Reassembly of Split Green Fluorescent
Protein. Journal of the American Chemical Society, 2011. 133(11): p. 4046-4052.
270. Talley, C.E., et al., Surface-Enhanced Raman Scattering from Individual Au
Nanoparticles and Nanoparticle Dimer Substrates. Nano Letters, 2005. 5(8): p. 1569-
1574.
196

271. Pinaud, F., et al., Bioactivation and Cell Targeting of Semiconductor CdSe/ZnS
Nanocrystals with Phytochelatin-Related Peptides. Journal of the American Chemical
Society, 2004. 126(19): p. 6115-6123.
272. Liu, Y., et al., Synthesis, stability, and cellular internalization of gold nanoparticles
containing mixed peptide-poly(ethylene glycol) monolayers. Anal Chem, 2007. 79(6): p.
2221-9.
273. Meakin, P., Simulation of the Kinetics of Aggregation - Fractal and Scaling. Croatica
Chemica Acta, 1992. 65(2): p. 237-267.
274. González, A.E., Universality of colloid aggregation in the reaction limit: The computer
simulations. Physical Review Letters, 1993. 71(14): p. 2248-2251.
275. Xie, Z., et al., Pure optical photoacoustic microscopy. Optics Express, 2011. 19(10): p.
9027-9034.
276. Kneipp, K., H. Kneipp, and J. Kneipp, Surface-Enhanced Raman Scattering in Local
Optical Fields of Silver and Gold NanoaggregatesFrom Single-Molecule Raman
Spectroscopy to Ultrasensitive Probing in Live Cells. Accounts of Chemical Research,
2006. 39(7): p. 443-450.
277. Hrelescu, C., et al., Single gold nanostars enhance Raman scattering. Applied Physics
Letters, 2009. 94(15): p. 153113.
278. Barbosa, S., et al., Tuning Size and Sensing Properties in Colloidal Gold Nanostars.
Langmuir, 2010. 26(18): p. 14943-14950.
279. Shim, M.G., et al., Study of Fiber-Optic Probes for in Vivo Medical Raman Spectroscopy.
Applied Spectroscopy, 1999. 53(6): p. 619-627.
280. Freudiger, C.W., et al., Label-Free Biomedical Imaging with High Sensitivity by
Stimulated Raman Scattering Microscopy. Science, 2008. 322(5909): p. 1857-1861.
281. Carrabba, M.M., et al., The Utilization of a Holographic Bragg Diffraction Filter for
Rayleigh Line Rejection in Raman Spectroscopy. Applied Spectroscopy, 1990. 44(9): p.
1558-1561.
282. Pelletier, M.J. and R.C. Reeder, Characterization of Holographic Band-Reject Filters
Designed for Raman Spectroscopy. Applied Spectroscopy, 1991. 45(5): p. 765-770.
283. Yang, B., M.D. Morris, and H. Owen, Holographic Notch Filter for Low-Wavenumber
Stokes and Anti-Stokes Raman Spectroscopy. Applied Spectroscopy, 1991. 45(9): p.
1533-1536.
284. Miyamoto, D., et al., Completely Dispersible PEGylated Gold Nanoparticles under
Physiological Conditions: Modification of Gold Nanoparticles with Precisely Controlled
PEG-b-polyamine. Langmuir, 2008. 24(9): p. 5010-5017.
285. Chen, G., et al., Measuring Ensemble-Averaged Surface-Enhanced Raman Scattering in
the Hotspots of Colloidal Nanoparticle Dimers and Trimers. Journal of the American
Chemical Society, 2010. 132(11): p. 3644-3645.
286. Bonham, A.J., et al., Detection of Sequence-Specific Protein-DNA Interactions via
Surface Enhanced Resonance Raman Scattering. Journal of the American Chemical
Society, 2007. 129(47): p. 14572-14573.
287. Perumal, J., et al., Design and fabrication of random silver films as substrate for SERS
based nano-stress sensing of proteins. RSC Advances, 2014. 4(25): p. 12995-13000.
288. Li, M., et al., Multiplexed detection of serological cancer markers with plasmon-
enhanced Raman spectro-immunoassay. Chemical science (Royal Society of Chemistry :
2010), 2015. 6(7): p. 3906-3914.
289. Dumortier, H., et al., Functionalized Carbon Nanotubes Are Non-Cytotoxic and Preserve
the Functionality of Primary Immune Cells. Nano Letters, 2006. 6(7): p. 1522-1528.
197

290. Cheng, K., et al., Construction and Validation of Nano Gold Tripods for Molecular
Imaging of Living Subjects. Journal of the American Chemical Society, 2014. 136(9): p.
3560-3571.
291. Loo, C., et al., Immunotargeted Nanoshells for Integrated Cancer Imaging and Therapy.
Nano Letters, 2005. 5(4): p. 709-711.
292. Feng, J., et al., Bioconjugation of Gold Nanobipyramids for SERS Detection and
Targeted Photothermal Therapy in Breast Cancer. ACS Biomaterials Science &
Engineering, 2017. 3(4): p. 608-618.

Abstract (if available)

Abstract In this thesis, a variety of novel split-fluorescent proteins (split-FPs) were developed and employed to design new surface enhanced Raman scattering (SERS) imaging nanoprobes capable of self-assembling into photonically active hot spots directly in live cells. These original optical probes use complementary split-FP fragments as molecular glue to control the assembly of gold or silver metal nanoparticles (AuNP or AgNP) clusters, whose tunable electronic and plasmonic properties have gained increasing interests for cellular and in vivo imaging applications. By appending complementary split-FP fragments at the surface of AuNPs or AgNPs, full-length and mature FPs can be reconstructed at the interface between clustered NPs and uniform plasmonic hot spots, a few nanometer in size, are generated. These hot spots provide massive near-field electromagnetic enhancements of the FP chromophore peculiar Raman fingerprints, allowing highly specific and sensitive SERS imaging of targeted cells. Because both complementary split-FP fragments are required to induce chromophore cyclization and activation of the FP Raman fingerprints, the detection selectivity of targeted cells by SERS imaging is significantly increased. Furthermore, FP-driven in situ metal nanoassemblies yield strong photoacoustic signal, allowing the FP/NP hybrid nanoclusters to serve as promising contrast agents for multimodal SERS and photoacoustic microscopy imaging with single cell sensitivity. ❧ A comprehensive background on the different components of this thesis work is provided in the introductory CHAPTER 1 together with more detailed explanations available in Annex sections. In the first part of CHAPTER 1, physical models of Raman spectroscopy and fundamentals of SERS are discussed with an emphasis on the advantages of SERS techniques. In the second part, we focus on the surface chemistry of metal nanoparticles, on the surface functional groups related to this project, and on the Raman reporters and their prominence for hot spot formation. To finish CHAPTER 1, we discuss the use of SERS nanoprobes as contrast agents for biomedical imaging and therapeutic/theranostic applications. CHAPTER 2 describes the development and the thorough characterization of novel split-FPs that we have used as Raman reporters in this thesis work. A quantitative analysis of the split-FP self-assembly process is presented through a comparative study of multiple spectral variants. The details of this self-assembly process hold great importance for this project because in addition to being Raman reporters upon complementation, split-FP fragments serve as complementary scaffolds to drive the formation of nanoassemblies. In CHAPTER 3, the surface functionalization of metal nanoparticles with complementary split-FP fragments and their assembly on targeted cells in vitro are described. A comprehensive characterization of the FP/NP hybrid probes by transmission electron microscopy analysis, dynamic light scattering technique, gel electrophoresis, and Raman spectroscopy is presented. Furthermore, cellular application of the SERS probes is studied by scanning electron microscopy analysis, SERS imaging and photoacoustic microscopy. Finally, in 0, future directions are proposed, such as the influence of nanomaterial geometry and the potential alterations of surface functional groups for dual targeting of different cell plasma membrane receptors. In this final chapter, prospective ideas to integrate different imaging techniques to improve in vivo experiment are also presented.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Electromagnetic modeling of plasmonic nanostructures

PDF

Monitoring local electric fields, charge densities at electrode surfaces using in-situ graphene/surface enhanced Raman spectroscopy (GERS/SERS) based Stark-shifts

PDF

Development of a quantitative proteomic molecular imaging platform using multiplexed nanoscale Raman scattering contrast agents

PDF

Surface plasmon resonant enhancement of photocatalysis

PDF

Flipping the switch on protein activity activity: elastin-like polypeptides assemble into cell switches and vesicles

PDF

Fabrication and application of plasmonic nanostructures: a story of nano-fingers

Asset Metadata

Creator Kӧker, Fatma Tuğba (author), Koker, Fatma Tugba (author)

Core Title Targeted cell imaging by in-situ assembly and activation of hot spot SERS nanoprobes using split-fluorescent protein scaffolds

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Molecular Biology

Publication Date 07/17/2017

Defense Date 06/22/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cell imaging,gold nanoparticles,hot spot,OAI-PMH Harvest,plasmonic,Raman,SERS imaging,silver nanoparticles,spectroscopy,split-fluorescent protein

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Pinaud, Fabien (committee chair), Cronin, Stephen (committee member), Fraser, Scott (committee member), Qin, Peter (committee member)

Creator Email fkoker@usc.edu,tugbakoker@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-400544

Unique identifier UC11264448

Identifier etd-KkerTuba-5522.pdf (filename),usctheses-c40-400544 (legacy record id)

Legacy Identifier etd-KkerTuba-5522.pdf

Dmrecord 400544

Document Type Dissertation

Rights Kӧker, Fatma Tuğba; Koker, Fatma Tugba

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA